article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
three dimensional quantum confinement of electrons ( _ e _ ) and holes ( _ h _ ) , such as in a semiconductor quantum dot ( qd ) potential , results in a variety of effects resulting from a discrete energy levels structure and an increased overlap between the _ e _ and _ h_. typically , a qd and a surrounding barrier are made of materials sharing a common cation , to recall example of inas / gaas , cdte / znte or cdse / znse qds .
the qds for which the qd and the barrier materials have no common atom are by far less investigated , despite of several advantages they offer . in the case of such qds with type - i confinement
one should mention here larger band offsets providing higher confining potentials for both , _ e _ and _ h_. such mixed structures may lead also to type - ii confinement , with a decreased wavefunction overlap of the confined _ e _ and _ h_. such increased _ e - h _ spatial separation and resulting enhancement of confined carriers lifetimes is highly attractive for studies of magnetic polaron in semimagnetic qds@xcite or aharonov - bohm effect,@xcite as well as for implementation of qds in photovoltaic devices@xcite .
the intermediate case between type - i and type - ii confinement , as in the present study , offers a chance of investigation of the impact of _ e _ and _ h _ wavefunctions overlap on properties of excitonic qds emission . in the present work , we design , produce and study individual cd(se , te ) qds embedded in znse barrier .
any composition gradient within the cd(se , te ) qd volume results in a decreased spatial overlap of wavefunctions of the _ e _ and _ h _ , due to localization of the _ e _ and _ h _ in regions , which are richer , respectively , in cdte and cdse ( see figure [ fig : bands ] providing a scheme of respective band alignments ) .
we implement a mn - assisted epitaxial growth , which allows us to obtain a very low spectral density of the qds , which enables us to study emission of single qds .
we find a wide distribution of exciton ( x ) and biexciton ( xx ) emission energy difference ( 3 mev @xmath0 @xmath1 @xmath0 26 mev ) .
a large scatter is found also for the x lifetime ( 280 - 620 ps ) .
a statistics collected on several qds indicates that the @xmath1 is strongly correlated with the x lifetime and a bright - dark exciton splitting @xmath3 , both quantities being a direct measure of _ e - h _ wavefunctions overlap.@xcite in addition , we establish that the @xmath1 is strongly correlated with such qd emission properties as the magnitude of the x fine structure exchange splitting @xmath2 and land - factor .
this indicates a way for a straightforward identification of qds with properties desired for a given implementation@xcite without a need of in - depth studies , like emission dynamics or magnetophotoluminescence .
the reported distinct dependence of the x fine structure splittings and land factor on a degree of _ e - h _ wavefunction overlap has been rarely observed in ii - vi qd systems so far , despite that it was well established in the case of iii - v low dimensional heterostructures@xcite .
here , we benefit from the fact that , in contrary to the case of typical binary ii - vi qd systems , the _ e - h _ separation changes significantly from qd to qd , enabling us to perform a systematic study .
for micro - photoluminescence ( @xmath4-pl ) studies the sample is placed in a pumped helium cryostat ( temperature down to @xmath5 k ) equipped with a superconducting split coil magnet
. a magnetic field of up to @xmath6 t is applied either in faraday ( @xmath7 ) or voigt ( @xmath8 ) configuration .
the magnetic field is perpendicular or parallel to the sample surface in this two respective cases .
the qds emission is excited above the barrier bandgap either in a continuous - wave mode at e@xmath9 = 3.06 ev ( @xmath10 = 405 nm ) or in a pulsed mode ( 100 fs pulse duration , 75 mhz repetition rate ) at 3.26 ev ( @xmath10 = 380 nm ) .
the excitation beam is focused on the sample surface down to 0.5 @xmath4 m diameter spot with an immersion type microscope objective@xcite ( na = 0.72 ) . in the time - integrated measurement
the signal is detected using a ccd camera coupled to a 0.75 m spectrometer ( 50 @xmath4ev of overall spectral resolution ) .
linear and circular polarizations of the signal are resolved . in the time - resolved @xmath4-pl measurements ,
the signal is detected by a hamamatsu photonics c5680 streak camera with 5 ps temporal resolution .
the scanning transmission electron microscope ( stem ) micrographs of the samples are obtained in high - angle annular dark field detector using fei tecnai osiris instrument .
the analysis is carried out at an acceleration voltage of 200 kev and point resolution of @xmath111.36 . in order to determine the concentration of zn , se , te and cd in the structure cross - section ,
a local chemical analysis by energy - dispersive x - ray spectroscopy ( edx ) is performed on the 100 nm thick specimen .
the tem specimen is prepared using the classical method of mechanical polishing followed by ion milling down to electron transparency .
so far , mixed , se and te based , qds were achieved primarily by a colloidal synthesis@xcite .
however , integration of such colloidal qds into operational devices might be challenging .
it should be easier in the case of epitaxially grown structures , which offer also a feasibility of p- and n- type doping .
on the other hand , strain resulting from a very large lattice mismatch@xcite between cdte ( a@xmath12 = 0.648 nm ) and znse ( a@xmath13 = 0.567 nm ) ( above 14% ) leads to a high spatial density of qds combining these two compounds .
this has precluded spectroscopy of individual qds in such mixed type systems so far.@xcite cd(se , te)/znse qds presented in this letter are grown by molecular beam epitaxy on a 1 @xmath4 m thick znse buffer deposited on a gaas substrate .
they form in a self - assembled mode out of atomic layers of se , cd , te , cd , and se ( each one for time of 15 sec ) consecutively deposited at temperature of the substrate set to 335 @xmath14c . an approximately monolayer thick cdse layer ( a@xmath15 = 0.608 nm ) inserted between cdte layer and znse diminishes the strain resulting from the cdte and the znse lattice mismatch.@xcite a small flux of the mn ions assists deposition of the te layer .
the qd layer is covered with a 80 nm thick znse cap . a reference sample , for which no mn ions are introduced is also grown under the same conditions .
the presence of mn induces a significant decrease of spectral density of qd emission lines ( see fig .
[ fig : gestosc_vs_mn ] ) . a spatial density of emitting qds estimated from @xmath4-pl spectra drops from 2000/@xmath4m@xmath16 in the case of the reference sample down to 100/@xmath4m@xmath16 in the case of the sample produced by the mn - assited growth .
the reduction of qd density for over an order of magnitude results most likely from a reduced mobility of adatoms on the sample surface during the qd nucleation when the mn atoms are present.@xcite the energy of qds emission lowered with respect to the reference sample case ( fig .
[ fig : gestosc_vs_mn ] ) suggests that the presence of mn induces also an increased qd size . a qualitatively similar , strong effect of mn incorporation on the surface morphology and plastic relaxation has been recently reported on also for
( al , ga)n / gan heterostructures.@xcite the haadf micrographs confirm the presence of a qd layer in the studied structures ( see fig .
[ fig : haadf]a ) ) . chemical analysis involving the edx spectroscopy reveals an expected decrease of the signal related to zn ( fig .
[ fig : haadf]b ) and an increase of the signal related to cd ( fig .
[ fig : haadf]c ) ) in the qd layer .
not complete vanishing of the signal related to the zn in the qd layer ( fig .
[ fig : haadf]b ) ) is understood when taking into account that the specimen thickness ( 100 nm ) is larger than an individual qd diameter .
in such a case signals from both , the cd(se , te ) qds and the znse barrier , contribute to the edx spectra in the qd layer spatial region .
figure [ fig : haadf]d ) shows the vertically integrated edx signal for zn , se , cd , te , mn as a function of the distance from the qds plane .
it confirms in a direct way the intended presence of the se and te atoms in the qds layer .
the average tellurium content in the qds is determined to @xmath17 using a procedure described in the appendix .
note that the mn concentration is too low to be detected by the chemical analysis .
a temporal evolution of the emission spectrum of a selected cd(se , te)/znse qd following the excitation pulse is presented in an image acquired by the streak camera ( fig .
[ fig : tau]a ) ) .
excitonic transitions shown are identified as the recombination of the x and xx confined in the same qd.@xcite a doublet of lines evidenced in the case of both , the x and xx , reflects a fine structure exchange splitting @xmath2 characteristic for a neutral exciton confined in an anisotropic qd.@xcite emission lifetime at the energy of the x and xx transitions is determined by fitting of a monoexponential decay curve to the emission intensity at respective cross - sections of the streak camera image ( see fig .
[ fig : tau]a ) ) .
the x lifetime determined for the selected qd equals to 450 ps . in the qd statistics ,
the x lifetime ranges from 620 ps , which is rather long for ii - vi epitaxial qds , down to 280 ps , in consistency with the previous reports on cdse / znse qds.@xcite the resulting decay rates @xmath18 determined as an inverse of the lifetime are plotted against the x - xx energy difference @xmath1 in fig .
[ fig : tau]b ) .
it is seen that the @xmath18 significantly increases with the increasing @xmath1 .
we note that the biexciton decay time of 120 ps corresponds to the decay rate ( @xmath19 ) around 9 ns@xmath20 , that is up to 6 times faster than the @xmath18 , much more than observed previously for the pure cdse / znse qds.@xcite the @xmath19 only slightly increases with the @xmath1 ( not shown ) . in order to asses factors governing the x lifetime , we determine the bright - dark exciton splitting @xmath3 resulting from isotropic component of the _ e - h _ exchange interaction .
the @xmath3 has been shown to be a direct measure of the _ e - h _
wavefunctions overlap.@xcite to determine the @xmath3 we brighten the dark exciton state x@xmath21 by application of the magnetic field in a direction parallel to the sample plane ( see the inset to fig . [
fig : tau]c)).@xcite we trace the energy position of the x@xmath21 as a function of the magnetic field and we determine its energy in the limit of zero magnetic field by extrapolation .
values of the @xmath3 of the order of 1 mev are found , as expected for ii - vi qds.@xcite we find that the @xmath3 increases with the increasing @xmath1 ( see fig .
[ fig : tau]c ) ) .
comparison of figs .
[ fig : tau]b ) and c ) reveals a high degree of correlation between the exciton radiative decay rate and the @xmath3 .
this confirms that : ( i ) dominating contribution to the @xmath3 comes from a short - range component of the _ e - h _ exchange interaction within the exciton , in agreement with previous theoretical considerations,@xcite ( ii ) the x decay in the studied qds is governed by a radiative recombination , while non - radiative processes are negligible .
a large range of variation of the @xmath1 , @xmath18 and @xmath3 suggests that the _ e - h _ spatial separation changes strongly from dot to dot .
the increase of the exciton radiative decay rate and of the @xmath3 with the increasing @xmath1 allows us to treat the @xmath1 as an estimate of the degree of _ e - h _ wavefunctions overlap in the further discussion .
time integrated measurements presented in sec .
[ sec : intpl ] shed more light on optical properties of the studied qds .
the figure [ fig : exampleqds ] shows example emission spectra of two individual qds ( labeled as qd1 and qd2 ) , both coming from the sample produced by the mn - assited growth .
the emission lines are identified as the x and xx recombination.@xcite the energy difference @xmath1 amounts to 2.99 mev and 20.38 mev in the case of the qd1 and the qd2 , respectively . as discussed in the previous section , the @xmath1 acts as a measure of the _ e - h _ wavefunction overlap .
thus , the qd1 and qd2 represent two extreme cases of a `` small '' and `` large '' _ e - h _ overlap , respectively .
note that the qd1 exhibits much lower intensity as compared to the qd2 , as expected for a qd with a decreased exciton oscillator strength resulting from a decreased _ e - h _
overlap.@xcite the spectra of the xx and x transitions from the qd1 and the qd2 plotted as a function of a detected linear polarization angle reveal two linearly polarized orthogonal components ( see panels c - f ) in the fig .
[ fig : exampleqds ] ) .
an exact antiphase of two components of the x doublet with respect to the xx doublet confirms that the transitions indeed originate from the same qd.@xcite the fine structure splitting @xmath2 for the x and xx from the qd1 amounts to 0.61 mev . in the case of the qd2 , much smaller @xmath2 equal to 0.07
mev is found .
+ figures .
[ fig : exampleqds]g ) and h ) show evolution of the x transition in the magnetic field applied in the faraday configuration .
the black lines represent the simulation of a zeeman splitting of the exciton ground state in an anisotropic qd following the equation @xmath22 , where @xmath23 represent transition energy at @xmath24 t , @xmath2 is the x fine - structure splitting , while @xmath25 and @xmath26 represent parallel excitonic land factor and bohr magneton , respectively .
the g - factor determined from the fit equals to 1.61 for the qd2 , which is quite typical value for cdse / znse qds .
in contrast , the @xmath27 found for the qd1 is atypically small as for the qd based on any selenium or tellurium compound .
the parameters describing the x emission determined in this subsection for the qd1 and qd2 suggest that in the studied qds the @xmath2 decreases , while the @xmath25 increases with the increasing @xmath1 .
the results obtained on a statistics of around twenty qds are discussed in the next subsection .
the values of the excitonic land factor and the @xmath2 determined for a number of cd(se , te)/znse qds are plotted against the @xmath1 in fig .
[ fig : stats]a ) .
the values for typical , binary type cdte / znte and cdse / znse qds are also presented for a reference ( after ref . ) .
a striking property of the studied qds is that the @xmath1 covers a very wide energy range from 3 mev to 25 mev ( see fig .
[ fig : tau ] and fig .
[ fig : stats ] ) . in the case of cdse or cdte
based qds the @xmath1 exhibits typically a low scatter and remains in a relatively narrow range of 20 - 25 mev@xcite or 12 - 15 mev@xcite , respectively .
hence , the @xmath1 in our case spans from values typical for the cdse / znse qds , through values typical for the cdte / znte qds , down to much lower values , which were not observed before for the cdse or the cdte based qds systems and which would not be expected for the cd(se , te)/znse qds if they were homogeneous in a volume .
the strongly reduced values of the @xmath1 mean that the _ e _ and _ h _ wavefunctions are separated in space much more than in pure cdse or cdte based qds .
it results most likely from a localization of the electrons in cdse and the holes in cdte rich regions ( see fig .
[ fig : bands ] ) present in the qd volume .
we link the presence of such regions to the qd layer growth procedure ( sec . [
sec : samples ] ) , namely the formation of the qds out of the deposited consecutively cdse and cdte layers .
a possible strain would widen the cdte bandgap and narrow the znse bandgap , even further enhancing the carrier localization effects.@xcite moreover , the fig . [
fig : stats ] reveals a clear increase of the excitonic land factor and a decrease of the @xmath2 with the increasing @xmath1 .
such dependencies have not been reported for epitaxial ii - vi qds so far . as it is seen in fig .
[ fig : stats]a ) , the g - factor attains the value as small as 0.95 for the qd characterized by the @xmath1 = 2.99 mev . at the first sight it is surprising ,
since typical g - factors of cdse and cdte qds are both equal or larger than 1.6 , what would result in the g - factor value exceeding 1.6 for a qd made of fully mixed ( cd , se)te .
this is apparently not the case , which gives a hint that the decreased @xmath1 indeed reflects the increased _ e - h _ separation resulting from a presence of cdte and cdse rich regions in the studied qds . for the cd(se , te)/znse qds with large values of the @xmath1 ,
the excitonic g - factor is approximately the same as for the pure cdse / znse qds ( see fig .
[ fig : stats]a ) ) , that is around 1.6.@xcite the increase of the parallel land factor with the decreasing size of the confining potential in the growth direction , was previously observed in the case of qws@xcite and arsenide based qds@xcite , as well as colloidal pbs nanocrystals.@xcite the effect was attributed to a quenching of the orbital momentum of a heavy hole when the hole wavefunction undergoes squeezing in the direction parallel to the magnetic field.@xcite the @xmath2 varies with the increasing @xmath1 from around 1 mev , the value larger than for reference cdse / znse or cdte / znte qds , down to the values below 0.06 mev ( see fig . [ fig : stats]b ) ) .
it has been shown previously that a dominant contribution to the @xmath2 comes from the long - range component of the _ e - h _
exchange interaction.@xcite the magnitude of the @xmath2 , similarly as the magnitude of the short - range component , have been shown to increase with a decreasing qd size ( thus with the increasing _ e - h _
overlap).@xcite thus , in the presently studied qds , one should expect enhancement of the @xmath2 with the @xmath1 .
this is , however , not the case , what indicates that in the studied qds , not just the overall _ e _ and _ h _ wavefunctions overlap , but specifically the anisotropy of the exciton wavefunction in the qd plane provides a main contribution to the @xmath2 .
a decrease of @xmath2 with a decreasing qd size was previously observed in the case of epitaxially grown inas / gaas qds @xcite and attributed to the dominating influence of piezoelectricity or heavy hole - light hole subbands mixing effects . in our case ,
the increase of the @xmath2 is correlated rather with the increasing te content in the cd(se , te)/znse qds , which translates into a larger respective spatial shift of the _ e _ and _ h _ in the qd plane .
we note that for the cd(se , te)/znse qds with large values of the @xmath1 not only the g - factor , but also the fine structure splitting @xmath2 ( 0.1 - 0.3 mev ) matches the one of the pure cdse / znse qds@xcite ( see fig . [
fig : stats]b ) ) .
this points toward binary type qd composition in the case of these qds .
this results most likely from an efficient substitution of te atoms by se atoms during the qd growth , due to a much higher sticking coefficient of the se atoms as compared to the te atoms.@xcite no systematic dependence of a diamagnetic coefficient @xmath28 on the @xmath1 is found ( not shown ) .
the @xmath28 is typically related to a span of the _ e _ and _ h _ wavefunctions in the qd plane .
thus , we state the _ e - h _ wavefunctions overlap varies from dot to dot mainly in the sample growth direction .
we have investigated systematically the impact of the electron - hole separation on optical properties of individual cd(se , te)/znse quantum dots .
the reduced spectral density of qd emission lines obtained thanks to the mn - assisted epitaxial growth has provided us an access to individual qds emission . a statistics involving several individual qds is collected .
we have found that there exists a high degree of correlation between quantities being a direct measure of the _ e - h _ wavefunctions overlap , such as the exciton radiative decay rate and the bright - dark exciton splitting @xmath3 , with the exciton - biexciton energy difference @xmath1 .
the @xmath1 varies significantly from dot to dot , i.e. , it changes from values typical for cdse / znse qds and cdte / znte qds , down to values as small as 3 mev , which were not reported previously for either of the two binary type systems .
we attribute the wide distribution of the @xmath1 to a wide range of variation of the studied qds composition and inhomogeneity affecting the _ e - h _ wavefunctions overlap . in particular ,
the strongly reduced @xmath1 points toward a presence of the cdte and cdse rich regions in the studied qds , which localize electron and hole , respectively , and account for an increased _ e - h _ separation . moreover , in contrary to a simple expectation , the cd(se , te)/znse qds g - factor is smaller than g - factors found previously in case of both , cdse / znse qds and cdte / znte , binary type qd systems
. additionally , a distinct and so far not observed for ii - vi qds , dependencies of the g - factor and of the fine structure exchange splitting @xmath2 on the @xmath1 are found .
the g - factor increases , while the @xmath2 decreases with the increasing @xmath1 , that is with the increasing _ e - h _ wavefunctions overlap .
since an ability of selection of qds with precisely defined emission properties is a prerequisite for qds practical implementations , the demonstrated parametrization of the g - factor or the @xmath2 by the @xmath1 in the studied qds provides a perspective for overcoming of the limits related to a scatter of the g - factor and the @xmath2 values observed typically in the qds ensemble . at the same time
, further efforts for obtaining qds with a degree of separation of confined carriers controlled on the production stage are still desired .
this work was supported by the polish national research center projects dec-2013/10/e
/ st3/00215 , dec-2013/09/b / st3/02603 , dec-2011/02/a / st3/00131 , dec-2015/18/e / st3/00559 , and dec-2011/02/a / st3/00131 , by the polish ministry of science and higher education in years 2012 - 2016 as a research grant `` diamentowy grant '' and `` iuventus plus '' grant ip2014 034573 . one of us ( t.s . ) was supported by the foundation for polish science through the start programme .
the work was carried out with the use of cept , cezamat , and nltk infrastructures financed by the european union - the european regional development fund within the operational programme `` innovative economy '' . the use of the fei tecnai osiris tem instrument located at the facility for electron microscopy & sample preparation of the university of rzeszw is acknowledged .
determination of the average molar density @xmath29 of the te atoms in the cd(se , te ) qds bases on relative changes of a vertically integrated edx signal .
formation of pure cdte qds would result in equal relative ( with respect to the maximum determined for the barrier layer ) decrease of zn and se signal in the qds layer . as evidenced in fig .
[ fig : haadf]d ) , showing the vertically integrated edx signal for of zn , se , cd , te , and mn as a function of the distance from the qds plane , this is , however , not the case .
the different relative decrease of the zn and se signal in the qds layer confirms the presence of the se atoms in the qds . since
a number of the cd anions in the cd(se , te ) qds is equal to the sum of the , substituting each other , te and se cations , the ratio of a relative decrease @xmath30 of the se signal to a relative decrease @xmath31 of the zn signal ( substituted by cd ) gives the ratio of the te to cd atom densities , thus the average molar density @xmath29 of the te atoms in the cd(se , te ) qds : @xmath32 we note that the edx signal detection sensitivity depends in unknown way on a given element . as a result , the absolute values of edx signal for each element are also not known .
it is known , however , that the signals for se and zn in the znse barrier region attain their maximum values .
these values serve as a reference in determination of relative changes of the edx signal in the qd layer for respective elements . | cd(se , te ) quantum dots ( qd ) in znse barrier typically exhibit a very high spectral density , which precludes investigation of single dot photoluminescence .
we design , grow and study individual cd(se , te)/znse qds of low spectral density of emission lines achieved by implementation of a mn - assisted epitaxial growth .
we find an unusually large variation of exciton - biexciton energy difference ( 3 mev @xmath0 @xmath1 @xmath0 26 mev ) and of exciton radiative recombination rate in the statistics of qds .
we observe a strong correlation between the exciton - biexciton energy difference , exciton recombination rate , splitting between dark and bright exciton , and additionally the exciton fine structure splitting @xmath2 and land factor . above results
indicate that values of the @xmath2 and of the land factor in the studied qds are dictated primarily by the electron and hole respective spatial shift and wavefunctions overlap , which vary from dot to dot due to a different degree of localization of electrons and holes in , respectively , cdse and cdte rich qd regions . |
spectral clustering is an important clustering technique that has been extensively studied in the image processing , data mining , and machine learning communities ( @xcite ) .
it is considered superior to traditional clustering algorithms like @xmath0-means in terms of having deterministic polynomial - time solution , the ability to model arbitrary shaped clusters , and its equivalence to certain graph cut problems .
for example , spectral clustering is able to capture the underlying moon - shaped clusters as shown in fig .
[ fig : two_moon](b ) , whereas @xmath0-means would fail ( fig .
[ fig : two_moon](a ) ) .
the advantage of spectral clustering has also been validated by many real - world applications , such as image segmentation ( @xcite ) and mining social networks ( @xcite ) .
+ spectral clustering was originally proposed to address an unsupervised learning problem : the data instances are unlabeled , and all available information is encoded in the graph laplacian .
however , there are cases where unsupervised spectral clustering becomes insufficient .
using the same toy data , as shown in ( fig .
[ fig : two_moon](c ) ) , when the two moons are under - sampled , the clusters become so sparse that the separation of them becomes difficult . to help spectral clustering recover from an undesirable partition
, we can introduce side information in various forms , in either small or large amounts .
for example : 1 . *
pairwise constraints * : domain experts may explicitly assign constraints that state a pair of instances must be in the same cluster ( must - link , ml for short ) or that a pair of instances can not be in the same cluster ( cannot - link , cl for short ) . for instance , as shown in fig .
[ fig : two_moon](d ) , we assigned several ml ( solid lines ) and cl ( dashed lines ) constraints , then applied our constrained spectral clustering algorithm , which we will describe later . as a result
, the two moons were successfully recovered .
* partial labeling * : there can be labels on some of the instances , which are neither complete nor exhaustive .
we demonstrate in fig .
[ fig : uci_ari ] that even small amounts of labeled information can greatly improve clustering results when compared against the ground truth partition , as inferred by the labels .
* alternative weak distance metrics * : in some situations there may be more than one distance metrics available . for example , in table [ table : reuters ] and accompanying paragraphs we describe clustering documents using distance functions based on different languages ( features ) .
* transfer of knowledge * : in the context of transfer learning ( @xcite ) , if we treat the graph laplacian as the target domain , we could transfer knowledge from a different but related graph , which can be viewed as the source domain .
we discuss this direction in section [ sec : algorithm : transfer ] and [ sec : exp : transfer ] .
all the aforementioned side information can be transformed into pairwise ml and cl constraints , which could either be hard ( binary ) or soft ( degree of belief ) .
for example , if the side information comes from a source graph , we can construct pairwise constraints by assuming that the more similar two instance are in the source graph , the more likely they belong to the same cluster in the target graph . consequently the constraints should naturally be represented by a degree of belief , rather than a binary assertion . how to make use of these side information to improve clustering falls into the area of constrained clustering ( @xcite ) . in general , constrained clustering is a category of techniques that try to incorporate ml and cl constraints into existing clustering schemes .
it has been well studied on algorithms such as @xmath0-means clustering , mixture model , hierarchical clustering , and density - based clustering .
previous studies showed that satisfying all constraints at once ( @xcite ) , incrementally ( @xcite ) , or even pruning constraints ( @xcite ) is intractable .
furthermore , it was shown that algorithms that build set partitions incrementally ( such as @xmath0-means and em ) are prone to being over - constrained ( @xcite ) .
in contrast , incorporating constraints into spectral clustering is a promising direction since , unlike existing algorithms , all data instances are assigned simultaneously to clusters , even if the given constraints are inconsistent .
constrained spectral clustering is still a developing area .
previous work on this topic can be divided into two categories , based on how they enforce the constraints .
the first category ( @xcite ) directly manipulate the graph laplacian ( or equivalently , the affinity matrix ) according to the given constraints ; then unconstrained spectral clustering is applied on the modified graph laplacian .
the second category use constraints to restrict the feasible solution space ( @xcite ) .
existing methods in both categories share several limitations : * they are designed to handle only binary constraints .
however , as we have stated above , in many real - world applications , constraints are made available in the form of real - valued degree of belief , rather than a yes or no assertion . *
they aim to satisfy as many constraints as possible , which could lead to inflexibility in practice .
for example , the given set of constraints could be noisy , and satisfying some of the constraints could actually hurt the overall performance . also , it is reasonable to ignore a small portion of constraints in exchange for a clustering with much lower cost .
* they do not offer any natural interpretation of either the way that constraints are encoded or the implication of enforcing them . in this paper
, we study how to incorporate * large * amounts of pairwise constraints into spectral clustering , in a flexible manner that addresses the limitations of previous work .
then we show the practical benefits of our approach , including new applications previously not possible .
we extend beyond binary ml / cl constraints and propose a more flexible framework to accommodate general - type side information .
we allow the binary constraints to be relaxed to real - valued degree of belief that two data instances belong to the same cluster or two different clusters .
moreover , instead of trying to satisfy each and every constraint that has been given , we use a user - specified threshold to lower bound how well the given constraints must be satisfied .
therefore , * our method provides maximum flexibility in terms of both representing constraints and satisfying them*. this , in addition to handling large amounts of constraints , allows the encoding of new styles of information such as entire graphs and alternative distance metrics in their raw form without considering issues such as constraint inconsistencies and over - constraining .
our contributions are : * we propose a principled framework for constrained spectral clustering that can incorporate large amounts of both hard and soft constraints .
* we show how to enforce constraints in a flexible way : a user - specified threshold is introduced so that a limited amount of constraints can be ignored in exchange for lower clustering cost .
this allows incorporating side information in its raw form without considering issues such as inconsistency and over - constraining .
* we extend the objective function of unconstrained spectral clustering by encoding constraints explicitly and creating a novel constrained optimization problem .
thus our formulation naturally covers unconstrained spectral clustering as a special case .
* we show that our objective function can be turned into a generalized eigenvalue problem , which can be solved deterministically in polynomial time .
this is a major advantage over constrained @xmath0-means clustering , which produces non - deterministic solutions while being intractable even for @xmath1 ( @xcite ) .
* we interpret our formulation from both the graph cut perspective and the laplacian embedding perspective .
* we validate the effectiveness of our approach and its advantage over existing methods using standard benchmarks and new innovative applications such as transfer learning .
this paper is an extension of our previous work ( @xcite ) with the following additions : 1 ) we extend our algorithm from 2-way partition to @xmath0-way partition ( section [ sec : algorithm : k - way ] ) ; 2 ) we add a new geometric interpretation to our algorithm ( section [ sec : interpret : geo ] ) ; 3 ) we show how to apply our algorithm to a novel application ( section [ sec : algorithm : transfer ] ) , namely transfer learning , and test it with a real - world fmri dataset ( section [ sec : exp : transfer ] ) ; 4 ) we present a much more comprehensive experiment section with more tasks conducted on more datasets ( section [ sec : exp:2moon ] and [ sec : exp : reuters ] ) . the rest of the paper is organized as follows : in section [ sec : related ] we briefly survey previous work on constrained spectral clustering ; section [ sec : background ] provides preliminaries for spectral clustering ; in section [ sec : model ] we formally introduce our formulation for constrained spectral clustering and show how to solve it efficiently ; in section [ sec : interpret ] we interpret our objective from two different perspectives ; in section [ sec : algorithm ] we discuss the implementation of our algorithm and possible extensions ; we empirically evaluate our approach in section [ sec : exp ] ; section [ sec : conclusion ] concludes the paper .
constrained clustering is a category of methods that extend clustering from unsupervised setting to semi - supervised setting , where side information is available in the form of , or can be converted into , pairwise constraints .
a number of algorithms have been proposed on how to incorporate constraints into spectral clustering , which can be grouped into two categories .
the first category manipulates the graph laplacian directly .
@xcite proposed the spectral learning algorithm that sets the @xmath2-th entry of the affinity matrix to 1 if there is a ml between node @xmath3 and @xmath4 ; 0 for cl . a new graph laplacian
is then computed based on the modified affinity matrix . in ( @xcite ) , the constraints are encoded in the same way , but a random walk matrix is used instead of the normalized laplacian .
@xcite proposed to add both positive ( for ml ) and negative ( for cl ) penalties to the affinity matrix ( they then used kernel @xmath0-means , instead of spectral clustering , to find the partition based on the new kernel ) .
@xcite proposed to propagate the constraints in the affinity matrix . in @xcite ,
the graph laplacian is modified by combining the constraint matrix as a regularizer .
the limitation of these approaches is that there is no principled way to decide the weights of the constraints , and there is no guarantee that how well the give constraints will be satisfied .
the second category manipulates the eigenspace directly .
for example , the subspace trick introduced by @xcite alters the eigenspace which the cluster indicator vector is projected onto , based on the given constraints .
this technique was later extended in @xcite to accommodate inconsistent constraints .
@xcite encoded partial grouping information as a subspace projection .
@xcite enforced constraints by regularizing the spectral embedding .
this type of approaches usually strictly enforce given constraints .
as a result , the results are often over - constrained , which makes the algorithms sensitive to noise and inconsistencies in the constraint set .
moreover , it is non - trivial to extend these approaches to incorporate soft constraints .
in addition , @xcite proposed a spectral kernel design that combines multiple clustering tasks .
the learned kernel is constrained in such a way that the data distributions of any two tasks are as close as possible . their problem setting differs from ours because we aim to perform single - task clustering by using two ( disagreeing ) data sources .
@xcite showed how to incorporate pairwise constraints into a penalized matrix factorization framework .
their matrix approximation objective function , which is different from our normalized min - cut objective , is solved by an em - like algorithm .
we would like to stress that the pros and cons of spectral clustering as compared to other clustering schemes , such as @xmath0-means clustering , hierarchical clustering , etc .
, have been thoroughly studied and well established .
we do not claim that constrained spectral clustering is universally superior to other constrained clustering schemes .
the goal of this work is to provide a way to incorporate constraints into spectral clustering that is more flexible and principled as compared with existing constrained spectral clustering techniques .
in this paper we follow the standard graph model that is commonly used in the spectral clustering literature .
we reiterate some of the definitions and properties in this section , such as graph laplacian , normalized min - cut , eigendecomposition and so forth , to make this paper self - contained .
readers who are familiar with the materials can skip to our formulation in section [ sec : model ] .
important notations used throughout the rest of the paper are listed in table [ table : notation ] .
.table of notations [ cols="^,<",options="header " , ] finally we apply our algorithm to transfer learning on the resting - state fmri data .
an fmri scan of a person consists of a sequence of 3d images over time .
we can construct a graph from a given scan such that a node in the graph corresponds to a voxel in the image , and the edge weight between two nodes is ( the absolute value of ) the correlation between the two time sequences associated with the two voxels .
previous work has shown that by applying spectral clustering to the resting - state fmri we can find the substructures in the brain that are periodically and simultaneously activated over time in the resting state , which may indicate a network associated with certain functions ( @xcite ) .
one of the challenges of resting - state fmri analysis is instability .
noise can be easily introduced into the scan result , e.g. the subject moved his / her head during the scan , the subject was not at resting state ( actively thinking about things during the scan ) , etc .
consequently , the result of spectral clustering becomes instable .
if we apply spectral clustering to two fmri scans of _ the same _ person on two different days , the normalized min - cuts on the two different scans are so different that they provide little insight into the brain activity of the subject ( fig .
[ fig : fmri](a ) and ( b ) ) . to overcome this problem
, we use our formulation to transfer knowledge from scan 1 to scan 2 and get a constrained cut , as shown in fig .
[ fig : fmri](c ) .
this cut represents what the two scans agree on . the pattern captured by fig .
[ fig : fmri](c ) is actually the _ default mode network _ ( dmn ) , which is the network that is periodically activated at resting state ( fig .
[ fig : fmri](d ) shows the idealized dmn as specified by domain experts ) . + to further illustrate the practicability of our approach , we transfer the idealized dmn in fig . [
fig : fmri](d ) to a set of fmri scans of elderly subjects .
the dataset was collected and processed within the research program of the university of california at davis alzheimer s disease center ( ucd adc ) .
the subjects were categorized into two groups : those diagnosed with cognitive syndrome ( 20 individuals ) and those without cognitive syndrome ( 11 individuals ) . for each individual scan
, we encode the idealized dmn into a constraint matrix ( using the rbf kernel ) , and enforce the constraints onto the original fmri scan .
we then compute the cost of the constrained cut that is the most similar to the dmn .
if the cost of the constrained cut is high , it means there is great disagreement between the original graph and the given constraints ( the idealized dmn ) , and vice versa . in other words ,
the cost of the constrained cut can be interpreted as the cost of transferring the dmn to the particular fmri scan . in fig .
[ fig : fmri : str ] , we plot the costs of transferring the dmn to both subject groups .
we can clearly see that the costs of transferring the dmn to people without cognitive syndrome tend to be lower than to people with cognitive syndrome .
this conforms well to the observation made in a recent study that the dmn is often disrupted for people with the alzheimer s disease ( @xcite ) .
in this work we proposed a principled and flexible framework for constrained spectral clustering that can incorporate large amounts of both hard and soft constraints .
the flexibility of our framework lends itself to the use of all types of side information : pairwise constraints , partial labeling , alternative metrics , and transfer learning .
our formulation is a natural extension to unconstrained spectral clustering and can be solved efficiently using generalized eigendecomposition .
we demonstrated the effectiveness of our approach on a variety of datasets : the synthetic two - moon dataset , image segmentation , the uci benchmarks , the multilingual reuters documents , and resting - state fmri scans .
the comparison to existing techniques validated the advantage of our approach .
we gratefully acknowledge support of this research via onr grants n00014 - 09 - 1 - 0712 automated discovery and explanation of event behavior , n00014 - 11 - 1 - 0108 guided learning in dynamic environments and nsf grant nsf iis-0801528 knowledge enhanced clustering .
amini mr , usunier n , goutte c ( 2009 ) learning from multiple partially observed views - an application to multilingual text categorization . in : advances in neural information processing systems 22
( nips 2009 ) , pp 2836 davidson i , ravi ss , ester m ( 2007 ) efficient incremental constrained clustering . in : proceedings of the 13th acm sigkdd international conference on knowledge discovery and data mining ( kdd 2007 ) , pp 240249 de bie t , suykens jak , de moor b ( 2004 ) learning from general label constraints . in : structural , syntactic , and statistical pattern recognition , joint iapr international workshops ( sspr / spr 2004 ) , pp 671679 ji x , xu w ( 2006 ) document clustering with prior knowledge . in : proceedings of the 29th annual international acm sigir conference on research and development in information retrieval ( sigir 2006 ) , pp 405412 martin d , fowlkes c , tal d , malik j ( 2001 ) a database of human segmented natural images and its application to evaluating segmentation algorithms and measuring ecological statistics . in : proceedings of the 8th international conference on computer vision ( iccv 2001 ) , vol 2 , pp 416423 wang f , ding chq , li t ( 2009 ) integrated kl ( k - means - laplacian ) clustering : a new clustering approach by combining attribute data and pairwise relations . in : proceedings of the 9th siam international conference on data mining ( sdm 2009 ) , pp 3848 xu q , desjardins m , wagstaff k ( 2005 ) constrained spectral clustering under a local proximity structure assumption . in : proceedings of the 18th international florida artificial intelligence research society conference , pp 866867 | constrained clustering has been well - studied for algorithms such as @xmath0-means and hierarchical clustering .
however , how to satisfy many constraints in these algorithmic settings has been shown to be intractable .
one alternative to encode many constraints is to use spectral clustering , which remains a developing area . in this paper
, we propose a flexible framework for constrained spectral clustering .
in contrast to some previous efforts that implicitly encode must - link and cannot - link constraints by modifying the graph laplacian or constraining the underlying eigenspace , we present a more natural and principled formulation , which explicitly encodes the constraints as part of a constrained optimization problem .
our method offers several practical advantages : it can encode the degree of belief in must - link and cannot - link constraints ; it guarantees to lower - bound how well the given constraints are satisfied using a user - specified threshold ; it can be solved deterministically in polynomial time through generalized eigendecomposition .
furthermore , by inheriting the objective function from spectral clustering and encoding the constraints explicitly , much of the existing analysis of unconstrained spectral clustering techniques remains valid for our formulation .
we validate the effectiveness of our approach by empirical results on both artificial and real datasets .
we also demonstrate an innovative use of encoding large number of constraints : transfer learning via constraints . |
the semiclassical analysis has proven to be a very useful tool in the field of `` quantum chaos '' as well as in many other fields .
nevertheless , one should bear in mind , that it only _ approximates _ the true quantal solution .
thus , it is imperative to know what are the _ errors _ which are inherent to the semiclassical approximation , and whether they could be considered as sufficiently small for the problem at hand .
we shall focus our attention on one particular application of the semiclassical approximation : the calculation of the energy spectra of classically chaotic systems .
the analytical tool that is used for this purpose is the semiclassical gutzwiller trace formula @xcite which expresses the _ quantum _ spectral density in terms of _ classical _ quantities , and in particular the actions and stabilities of classical periodic orbits .
the trace formula was used , among other things , to explain and discuss spectral statistics and their relation to the universal predictions of random matrix theory ( rmt ) @xcite . however , a prerequisite for the use of the semiclassical approximation to compute short
range statistics is that it is able to reproduce the exact spectrum within an error which is comparable to or less than the mean level spacing !
this is a demanding requirement , and quite often , the ability of the semiclassical approximation to reproduce precise levels for high dimensional systems is doubted , and on the following grounds .
the mean level spacing depends on the dimensionality ( number of freedoms ) of the system , and it is @xmath6 .
gutzwiller @xcite uses an argument by pauli @xcite to show that in general the error margin for the semiclassical approximation scales as @xmath7 _ independently of the dimensionality_. applied to the trace formula , the expected error in units of the mean spacing , which is the figure of merit in the present context , is therefore expected to be @xmath8
. we shall refer to this as the traditional estimate " .
it sets @xmath9 as a critical dimension for the applicability of the semiclassical trace formula and hence for the validity of the conclusions which are drawn from it .
the few systems in @xmath10 dimensions which were numerically investigated display spectral statistics which adhere to the predictions of rmt as accurately as their counterparts in @xmath9 @xcite .
thus , the `` traditional estimate '' can not be entirely correct in the present context , and we shall explain the reasons why it is inadequate when we discuss other error bounds in the next section .
it is rather surprising that the problem of the accuracy of the semiclassical trace formula was rather rarely touched upon in the literature .
gutzwiller quotes the `` traditional '' estimate of @xmath8 which was discussed above @xcite .
gaspard and alonso @xcite and alonso and gaspard @xcite derived explicit ( generic ) @xmath11 corrections for the periodic orbit terms in the trace formula , but did not investigate their effect on the semiclassical accuracy of energy levels .
diffraction corrections were discussed in the context of the trace formula by vattay , wirzba and rosenqvist @xcite and by primack et al . @xcite . also in these works the focus is on the corrections to individual periodic orbit terms rather than on the overall effect on energy levels .
boasman @xcite studied the accuracy of the boundary integral method ( bim ) @xcite for 2d billiards in the case that the exact kernel is replaced by its semiclassical asymptotic approximation .
he finds that the resulting error is of the same magnitude as the mean spacing , which is consistent with the traditional estimate .
however , boasman s work does not refer directly to the trace formula and to the periodic orbits contributions .
the works of bleher @xcite and of prosen and robnik @xcite discuss the accuracy of the semiclassical approximation in the integrable case .
the purpose of this work is to address the subject of the accuracy of the semiclassical trace formula conceptually , theoretically and numerically .
we shall be interested in particular in the dependence of the semiclassical error on the _ dimension _ of the system .
to do so , we shall have to start by developing the basic concepts and define the measures we use for a quantitative estimate of the spectral error ( section [ sec : accuracy - measures ] ) .
the accuracy of the semiclassical approximation of the quantal energy spectrum will be then studied via the dual classical spectrum of actions and stabilities of periodic orbits ( time spectrum ) .
this will enable us to use our data base of quantum levels and periodic orbits for the sinai billiards in two and three dimensions for a direct evaluation of the semiclassical error ( section [ sec : sinai - numerics ] ) .
we shall summarize the paper and discuss a few relevant points in section [ sec : discussion ] .
in order to define a proper error measure for the semiclassical approximation of the energy spectrum one has to clarify a few issues .
in contrast with the ebk quantization which gives an _ explicit _
formula for the spectrum , the semiclassical spectrum for chaotic systems is _ implicit _ in the trace formula , or in the semiclassical expression for the spectral determinant . to extract the semiclassical spectrum we recall that the exact spectrum , @xmath12 , can be obtained from the exact counting function : @xmath13 by solving the equation @xmath14 in the last equation
, an arbitrarily small amount of smoothing must be applied to the heavyside function . in complete analogy ,
one obtains the semiclassical spectrum @xmath15 as @xcite : @xmath16 where @xmath17 is a semiclassical approximation of @xmath18 .
note that @xmath17 with which we start is not necessarily a sharp counting function .
however , once @xmath15 is known , we can `` rectify ''
the smooth @xmath17 into the sharp counting function @xmath19 @xcite : @xmath20 the simplest choice for @xmath17 is the gutzwiller trace formula @xcite truncated at the heisenberg time , which is what we shall use in the present paper .
alternatively , one can start from the regularized berry keating zeta function @xmath21 @xcite , and define @xmath22 , in which case @xmath23 .
next , in order to define a quantitative measure of the semiclassical error , one should establish a _ correspondence _ between the quantal and the semiclassical levels , namely , one should identify the semiclassical counterparts of the exact quantum levels . in classically chaotic systems , for which the gutzwiller trace formula is applicable , the only constant of the motion is the energy .
this is translated into a single `` good '' quantum number in the quantum spectrum , which is the ordinal number of the levels when ordered by their magnitude .
thus , the only correspondence which can be established between the exact spectrum @xmath24 and its semiclassical approximation , @xmath25 , is @xmath26 this is to be contrasted with integrable systems , where it is appropriate to compare the exact and approximate levels which have the same quantum numbers .
the scale on which the accuracy of semiclassical energy levels should be measured depends , in general , on the problem at hand .
the most natural choice , however , is the mean level spacing @xmath27 where @xmath28 is the smooth density of states .
the semiclassical error of the nth level is therefore measured by @xcite : @xmath29 a more useful and significant measure is the average of @xmath30 over an energy interval @xmath31 centered at @xmath32 , which contains a large number of quantum energies , but which must be so small that both the classical dynamics and the mean density of states remain approximately constant .
this energy averaging will be denoted by triangular brackets @xmath33 in the sequel .
if the semiclassical mean density @xmath34 agrees with @xmath28 to a high precision , then obviously @xmath35 . in this case
@xmath36 is not a useful error measure . for billiard systems
this is always the case , since the mean spectral density can be written as an asymptotic series with explicitly known coefficients @xcite . for general systems ,
only the leading weyl term is explicitly known .
two appropriate measures which are sensitive to the accuracy of the fluctuating parts of the level densities are the mean absolute error : @xmath37 and the variance : @xmath38 having defined the spectral error measures , let us apply them and try to get some estimates of the semiclassical error . in the introductory section we have mentioned the `` traditional '' estimate of the semiclassical error .
gutzwiller @xcite shows , based on pauli @xcite , that the inherent error in the semiclassical ( van - vleck ) approximation of the quantal time propagator scales like @xmath7 . since the energy levels are the temporal fourier components of the propagator , it is plausible to assume that they have the same degree of accuracy : @xmath39 ( strictly speaking , this is an upper bound . )
the mean density of energy levels , for a general @xmath2-dimensional system is asymptotically given by weyl s formula @xcite : @xmath40 where @xmath41 is the measure of the energy surface in the classical phase space .
hence , @xmath42 that is , the semiclassical approximation is ( marginally ) accurate in 2 dimensions , but it fails to predict accurate energy levels for 3 dimensions or more .
one may get a different estimate of the semiclassical error , if the gutzwiller trace formula ( gtf ) is used as a starting point .
suppose that we have calculated @xmath17 to a certain degree of precision , and we compute from it the semiclassical energies using ( [ eq : scqc ] ) .
the quality of this approximation can be estimated if the leading corrections @xmath43 are also included and the resulting energy differences @xmath44 are evaluated .
we thus need to consider : @xmath45 combining ( [ eq : scqc ] ) and ( [ eq : impscqc ] ) we get to first order in @xmath44 : @xmath46 in the above we assumed that the fluctuations of @xmath17 around its average are not very large .
thus , @xmath47 let us apply the above formula and consider the case in which we take for @xmath17 its mean part @xmath48 , and that we include in @xmath48 terms of order up to ( and including ) @xmath49 . for @xmath43
we use both the leading correction to @xmath48 and the leading order periodic orbit sum which is ( formally ) of order @xmath50 .
hence , @xmath51 we conclude , that approximating the energies only by the mean counting function @xmath48 up to ( and not including ) the constant term , is already sufficient to obtain semiclassical energies which are accurate to @xmath52 with respect to the mean density of states .
note again , that no periodic orbits were included in @xmath17 .
including less terms in @xmath48 will lead to a diverging semiclassical error , while more terms will be masked by the periodic orbit ( oscillatory ) term .
one can do even better if one includes in @xmath17 the smooth terms up to and including the constant term ( @xmath53 ) together with the leading order periodic orbit sum which is formally also @xmath53 .
the semiclassical error is then : @xmath54 that is , the semiclassical energies measured in units of the mean level spacing are asymptotically accurate independently of the dimension !
this estimate grossly contradicts the traditional " estimate ( [ eq : epstrad ] ) and calls for an explanation .
the first point that should be noted is that the order of magnitude ( power of @xmath11 ) of the periodic orbit sum , which we considered above to be @xmath53 , is only a formal one . indeed ,
each term which is due to a _
single _ periodic orbit is of order @xmath55 .
however the periodic orbit sum _ absolutely diverges _ , and at best it is only _ conditionally convergent_. to give it a numerical meaning , the periodic orbit sum must therefore be regularized .
this is effectively achieved by truncating the trace formula or the corresponding spectral @xmath56 function @xcite
. however , the truncation cutoff itself depends on @xmath11 .
one can conclude , that the simple minded estimate ( [ eq : epsgtf ] ) given above is at best a lower bound , and the error introduced by the periodic orbit sum must be re - evaluated with more care .
this point will be dealt with in great detail in the sequel , and we shall eventually develop a meaningful framework for evaluating the magnitude of the periodic orbit sum . the connection and the disparity between the traditional " estimate of the semiclassical error and the one based on the trace formula can be further illustrated by the following argument .
the periodic orbit formula is derived from the semiclassical propagator @xmath57 using further approximations @xcite .
one thus wonders , how can it be that _ further _ approximations of @xmath57 actually _ reduce _ the semiclassical error from ( [ eq : epstrad ] ) to ( [ eq : epsgtf ] ) ?
the puzzle is resolved if we recall , that in order to obtain @xmath58 above we separated the density of states into a smooth and an oscillating parts , and we required that the smooth part is accurate enough . to achieve this
, we have to go beyond the leading weyl s term and to use specialized methods to calculate the smooth density of states beyond the leading order .
these methods are mostly developed for billiards @xcite . in any case , to obtain @xmath58 we have added an _ additional _ information which goes beyond the leading semiclassical approximation .
a straightforward check of the accuracy of the semiclassical spectrum using the error measures @xmath59 is exceedingly difficult due to the large number of periodic orbits needed because of the exponential proliferation in chaotic systems .
the few cases where such tests were carried out involve 2d systems and it was possible to check only the lowest ( less than a hundred ) levels ( e.g. @xcite ) .
the good agreement between the exact and the semiclassical values confirmed the expectation that in 2d the semiclassical error is small . in 3d ,
the topological entropy is typically much larger @xcite , and the direct test of the semiclassical spectrum becomes prohibitive .
facing with this grim reality , we have to introduce alternative error measures which yield the desired information , but which are more appropriate for a practical calculation .
we construct the measure : @xmath60 as before , the triangular brackets indicate averaging over an energy interval @xmath31 about @xmath32 .
we shall now show that @xmath61 faithfully reflects the deviations between the spectra , and is closely related to @xmath62 and @xmath63 .
note , that the following arguments are purely statistical and apply to every pair of staircase functions .
suppose first , that all the differences @xmath64 are smaller than the mean spacing .
then , @xmath65 is either 0 or 1 ( see figure [ fig : delta2small ] ) and hence @xmath66 however , the right hand side of the above equation ( the fraction of non zero contributions ) equals @xmath62 .
thus , @xmath67 if , on the other hand , deviations are much larger than one mean spacing , the typical horizontal distance @xmath68 should be comparable to the vertical distance @xmath65 , and hence , in this limit @xmath69 therefore , we expect @xmath70 to interpolate between @xmath62 and @xmath71 throughout the entire range of deviations . this behavior is indeed observed in a numerical test which was performed to check the above expectations .
we considered the unfolded exact spectrum ( normalized to unity mean spacing @xcite ) of the 3d sinai billiard @xmath72 and created from it a synthetic spectrum by adding a random variable with @xmath73 mean and variance @xmath74 : @xmath75 the @xmath76 has also a unity mean spacing and it is meant to imitate a semiclassical spectrum . after sorting @xmath76 we calculated the measures @xmath77 and @xmath70 as functions of @xmath78 .
the results are shown in figure [ fig : delta2-eps ] , and they verify the estimates ( [ eq : delta2small ] , [ eq : delta2large ] ) in the appropriate limits . the numerical test reported in figure [ fig : delta2smootheps2 ] demonstrates another attractive feature of the measure @xmath61 : it is completely equivalent to @xmath71 when the spectral counting functions are replaced by their smooth counterparts , provided that the smoothing width is of the order of 1 mean level spacing and the same smoothing is applied to both counting functions .
that is , @xmath79 for all deviations .
( in fact , for small deviations there is a proportionality factor , but it can be set to 1 if an appropriate smoothing is used . ) in testing the semiclassical accuracy , this kind of smoothing is essential and will be introduced by truncating the trace formula at the heisenberg time @xmath80 .
these properties of the measure @xmath70 , and its complete equivalence to @xmath71 for smooth counting functions , renders it a most appropriate measure of the semiclassical error .
we now turn to the practical evaluation of @xmath70 . to affect the energy averaging , we choose a positive window function @xmath81 which has a width @xmath31 near @xmath32 and is normalized by @xmath82 .
it falls off sufficiently rapidly so that all the expressions which follow are well behaved .
construct the following counting functions that have an effective support on an interval of size @xmath31 about @xmath32 : @xmath83 at this stage @xmath84 and @xmath85 are still sharp staircases , and we note that the multiplication with @xmath86 preserves the sharpness of the stairs ( it is not a convolution ! ) .
we now explicitly construct @xmath87 as : @xmath88 to construct @xmath89 we need to smooth @xmath90 over a scale of order of one mean spacing .
this can be done e.g. by replacing the sharp stairs by error functions . as for @xmath91
, we prefer to simply replace it with the original @xmath17 , which we assume to be smooth over one mean spacing .
that is , we suppose that @xmath17 contains periodic orbits up to heisenberg time .
hence , @xmath92 a comment is in order here . strictly speaking , to satisfy ( [ eq : delta2epsilon2 ] ) we need to apply the same smoothing to @xmath18 and to @xmath91 , and in general @xmath93 , but there are differences of order 1 between the two functions .
however , since our goal is to determine whether the semiclassical error remains finite or diverges in the semiclassical limit @xmath94 , we disregard such inaccuracies of order 1 .
if more accurate error measure is needed , then more care should be practiced in this and in the following steps .
applying parseval s theorem to ( [ eq : delta2smoothint ] ) we get : @xmath95 where @xmath96 we shall refer to @xmath97 as the ( regularized ) quantal and semiclassical time spectra , respectively .
this name can be justified by invoking the gutzwiller trace formula and expressing the semiclassical counting function as a mean part plus a sum over periodic orbits .
we have : @xmath98 \ , \label{eq : gtf}\ ] ] where @xmath99 is the semiclassical amplitude of the jth periodic orbit , and @xmath100 are its period , action , maslov index , monodromy and repetition index , respectively .
then , the corresponding time spectrum reads : @xmath101 } \hat{w}([t + t_j(e)]/\hbar ) - \right .
\nonumber \\ & & \left .
\ ; \ ; \ ; \ ; \ ; \ ; \ ; \ ; \ ; \ ; \ ; \ ; \ ; \ ; \ ; \ ; \ ; \ ; \ ; \ ; \ ; \ ; \ ; { \rm e}^{(i/\hbar)[e t - s_j(e ) ] } \hat{w}([t - t_j(e)]/\hbar ) \right\ } .
\nonumber\end{aligned}\ ] ] in the above , the fourier transform of @xmath86 is denoted by @xmath102 .
it is a localized function of @xmath103 whose width is @xmath104 .
the sum over the periodic orbits in @xmath105 therefore produces sharp peaks centered at times that correspond to the periods @xmath106 , hence the name `` time spectrum '' .
the term @xmath107 corresponds to the smooth part and thus is sharply peaked near @xmath108 . to obtain ( [ eq : scltspt ] ) we expanded the actions near @xmath32 to first order : @xmath109 .
we note in passing , that this approximate expansion of @xmath110 can be avoided altogether if one performs the fourier transform over @xmath111 rather than over the energy .
this way , an action spectrum will emerge , but also here the action resolution will be finite , because the range of @xmath111 should be limited to the range where @xmath112 is approximately constant .
it turns out therefore , that the two approaches are essentially equivalent , and for billiards they are identical .
the manipulations done thus far were purely formal , and did not manifestly circumvent the difficult task of evaluating @xmath113 . however , the introduction of the time spectra and the formula ( [ eq : delta2ls ] ) put us in a better position than with the original expression ( [ eq : delta2int ] ) .
the advantages of using the time spectra in the present context are the following : * the semiclassical time spectrum @xmath114 is absolutely convergent for all times ( as long as the window function @xmath86 is well behaved , e.g. it is a gaussian ) . this statement is correct even if the sum ( [ eq : scltspt ] ) extends over the entire set of periodic orbits !
this is in contrast with the trace formula expression for @xmath17 ( and therefore @xmath115 ) which is absolutely divergent if all of the periodic orbits are included .
* time scale separation : as we noted above , the time spectrum is peaked at times that correspond to periods of the classical periodic orbits .
this allows us to distinguish between various qualitatively different types of contributions to @xmath89 .
we shall now pursue the separation of the time scales in detail .
we first note , that due to @xmath116 being real , there is a @xmath117 symmetry in ( [ eq : delta2ls ] ) and therefore @xmath118 .
we now divide the time axis into four intervals : @xmath119 : the shortest time scale in our problem is @xmath120 .
the contributions to this time interval are due to the differences between the exact and the semiclassical _ mean _ densities of states .
this is an important observation , since it allows us to distinguish between the two sources of semiclassical error the error that emerges from the mean densities and the error that originates from the fluctuating part ( periodic orbits ) .
since we are interested only in the semiclassical error that results from the fluctuating part of the spectral density , we shall ignore this regime in the following .
: this is the non universal regime @xcite , in which periodic orbits are still sparse , and can not be characterized statistically in a significant fashion .
the `` ergodic '' time scale @xmath121 is purely classical and is independent of @xmath11 .
@xmath122 : in this time regime periodic orbits are already in the universal regime and are dense enough to justify a statistical approach to their proliferation and stability .
the upper limit of this interval is the heisenberg time @xmath123 , which is the time that is needed to resolve the quantum ( discrete ) nature of a wavepacket with energy concentrated near @xmath32 .
the heisenberg time is `` quantum '' in the sense that it is dependent of @xmath11 : @xmath124 .
@xmath125 : this is the interval of `` long '' orbits which is effectively truncated from the integration as a result of introducing a smoothing of the quantal and semiclassical counting functions , with a smoothing scale of the order of a mean level spacing . dividing the integral ( [ eq : delta2ls ] ) according to the above time intervals ,
we can rewrite @xmath89 : @xmath126 as explained above , @xmath127 can be ignored due to smoothing on the scale of a mean level spacing .
the integral @xmath128 is to be neglected for the following reason .
the integral extends over a time interval which is finite and independent of @xmath11 , and therefore it contains a fixed number of periodic orbits contributions .
the semiclassical approximation provides , for each individual contribution , the leading order in @xmath11 , and therefore @xcite we should expect : @xmath129 the purpose of this work is to check whether the semiclassical error is finite or divergent as @xmath130 , and to study if the rate of divergence depends on dimensionality . equation ( [ eq : delta2-short ] ) implies that @xmath128 can not affect @xmath70 in the semiclassical limit and we shall neglect it in the following .
we thus remain with a lower bound for our measure : @xmath131 which is going to be our object of interest from now on . the fact that @xmath132 is extremely large on the classical scale
renders the calculation of all the periodic orbits with periods less than @xmath132 an impossible task .
however , sums over periodic orbits when the period is longer than @xmath121 tend to meaningful limits , and hence , we would like to recast the expression for @xmath133 in the following way .
write @xmath134 as : @xmath135 \label{eq : ct } \\
& \equiv & \frac{2}{\hbar } \int_{t_{\rm erg}}^{t_{\rm h } } { \rm d}t \ , \left\langle \left| \hat{d}(t ) \right|^2 \right\rangle_{t } \times c(t ) \\ & = & \int_{t_{\rm erg}}^{t_{\rm h } } \mbox{\large envelope } \times \mbox{\large correlation } \nonumber \,\end{aligned}\ ] ] where the parametric dependence on @xmath32 was omitted for brevity .
the smoothing over @xmath103 is explicitly indicated to emphasize that one may use a statistical interpretation of the terms of the integrand .
this is so because in this domain , the density of periodic orbits is so large , that within a time interval of width @xmath136 there are exponentially many orbits whose contributions are averaged due to the finite resolution .
we note now that we can use the following relation between the time spectrum and the spectral form factor @xmath137 : @xmath138 where @xmath139 is the scaled time .
the above form factor is smoothed according to the window function @xmath86 . hence : @xmath140 for generic chaotic systems we expect that @xmath137 agrees with the results of rmt in the universal regime @xmath141 @xcite , and therefore @xmath142 where @xmath143 for systems which violate time reversal symmetry , and @xmath144 if time reversal symmetry is respected .
this implies that the evaluation of @xmath89 reduces to @xmath145 the dependence on @xmath11 in this expression comes from the lower integration limit which is proportional to @xmath146 as well as from the implicit dependence of the function @xmath147 on @xmath11 .
formula ( [ eq : delta2final ] ) is our main theoretical result .
however , we do not know how to evaluate the correlation function @xmath148 from first principles .
the knowledge of the @xmath149 corrections to each of the terms in the semiclassical time spectrum is not sufficient since the resulting series which ought to be summed is not absolutely convergent ( see detailed discussion in section ( [ sec : discussion ] ) ) .
therefore we have to recourse to a numerical analysis , which will be described in the next section .
the numerical approach requires one further approximation , which is imposed by the fact that the number of periodic orbits with @xmath150 is prohibitively large .
we had to limit the data base of periodic orbit to the domain @xmath151 with @xmath152 .
the time @xmath153 has no physical origin , and it represents only the limits of our computational resources . using the available numerical data we were able to compute @xmath154 numerically for all @xmath155 and we then _ extrapolated _ it to the entire domain of interest .
we consider this extrapolation procedure to be the main source of uncertainty .
however , since the extrapolation is carried out in the _ universal _ regime , it should be valid if there are no other time scales between @xmath121 and @xmath132 .
we used the formalism and definitions presented above to check the accuracy of the semiclassical spectra of the 2d and 3d sinai billiards .
the most important ingredient in this numerical study is that we could apply the _ same _ analysis to the two systems , and by comparing them to give a reliable answer to the main question posed in this work , namely , how does the semiclassical accuracy depend on dimensionality .
the classical dynamics in billiards depends trivially on the energy ( velocity ) , and therefore the relevant parameter is the length rather than the period of the periodic orbits . because of the same reason , the quantum wavenumbers @xmath156 are the relevant variables in the quantum description . from now on we shall use the variables @xmath157 instead of @xmath158 , and use `` length spectra '' rather than `` time spectra '' .
the semiclassical limit is obtained for @xmath159 and @xmath160 is equivalent to @xmath161 .
note also that for a billiard @xmath162 where @xmath163 is a proportional to the billiard s volume .
the numerical work is based on the quantum spectra and on the classical periodic orbits which were calculated by schanz and smilansky @xcite for the 2d billiard , and by primack and smilansky @xcite for the 3d billiard .
the numerical methods and the checks performed to ensure that the quantum and the classical data bases are accurate , complete and immaculate are discussed in the papers cited above .
we start with the 2d sinai billiard , which is the free space between a square of edge @xmath164 and an inscribed disc of radius @xmath165 , with @xmath166 . in our case we used @xmath167 and @xmath168 and considered the quarter desymmetrized billiard ( see figure [ fig : sb2d - sb3d ] ) with dirichlet boundary conditions for the quantum calculations .
the quantal data base consisted of the lowest 27645 eigenvalues in the range @xmath169 , with eigenstates which are either symmetric or antisymmetric with respect to reflection on the main diagonal .
the classical data base consisted of the shortest 20273 periodic orbits ( including time reversal , reflection symmetries and repetitions ) in the length range @xmath170 . for each orbit ,
the length , the stability determinant and the reflection phase were recorded . [ cols="^,^,^ " , ] we begin the numerical analysis of the 2d sinai billiard by numerically demonstrating the correctness of equation ( [ eq : delta2-short ] ) .
that is , that for each _ individual _ periodic orbit , the semiclassical error indeed vanishes in the semiclassical limit . in figure
[ fig : sb2d - qm - sc ] we plot @xmath171 for @xmath172 as a function of @xmath173 .
this length corresponds to the shortest periodic orbit , that is , the one that runs along the edges that connect the circle with the outer square . for @xmath105 we used the gutzwiller trace formula .
as is clearly seen from the figure , the quantal
semiclassical difference indeed vanishes ( approximately as @xmath174 ) , in accordance with ( [ eq : delta2-short ] ) .
we emphasize again , that this behavior does not imply that @xmath70 vanishes in the semiclassical limit , since the number of terms depends on @xmath173 .
it implies _ only _ that @xmath128 vanishes in the limit , since it consists of a fixed and finite number of periodic orbit contributions .
we should also comment that ( non generic ) penumbra corrections to individual grazing orbits introduce errors which are of order @xmath175 with @xmath176 @xcite .
however , since the definition of `` grazing '' is in itself @xmath173 dependent , one can safely neglect penumbra corrections in estimating the large @xmath173 behavior of @xmath128 .
we now turn to the main body of the analysis , which is the evaluation of @xmath133 for the 2d sinai billiard . based on the available data sets , we plot in figure [ fig : dcxavg.2 ] the function @xmath177 in the interval @xmath178 for various values of @xmath173 .
one can observe , that as a function of @xmath179 the functions @xmath177 fluctuate in the interval for which numerical data were available , without exhibiting any systematic mean trend to increase or to decrease .
we therefore approximate @xmath177 by @xmath180 according to the discussion in section [ sec : accuracy - measures ] we extrapolate this formula in @xmath179 up to the heisenberg length @xmath181 and using ( [ eq : delta2final ] ) we obtain : @xmath182 the last equality is due to @xmath183 . to evaluate @xmath184 we averaged @xmath177 over the interval @xmath185 and
the results are shown in figure [ fig : dcavg.2 ] .
we choose @xmath186 because the density of periodic orbits is already large for this length ( see figure [ fig : dcxavg.2 ] ) and we expect universal behavior of the periodic orbits .
( for the sinai billiard described by flow the approach to the invariant measure is algebraic rather than exponential @xcite , and thus one can not have a well - defined @xmath187 . an any rate , the specific choice of @xmath187 did not affect the results in any appreciable way . ) inspecting @xmath184 , it is difficult to arrive at firm conclusions , since it seems to fluctuate around a constant value up to @xmath188 and then to decline .
if we approximate @xmath184 by a constant , we get a `` pessimistic '' value of @xmath70 : @xmath189 while if we assume that @xmath184 decays as a power - law , @xmath190 , then @xmath191 collecting the two bounds we get : @xmath192 our estimates for the 2d sinai billiard can be summarized by saying that the semiclassical error diverges no worse than logarithmically ( meaning , very mildly ) .
it may well happen that the semiclassical error is constant or even vanishes in the semiclassical limit . to reach a conclusive answer one should invest exponentially larger amount of numerical work .
there are a few comments in order here .
firstly , the quarter desymmetrization of the 2d sinai billiard does not exhaust its symmetry group , and in fact , a reflection symmetry around the diagonal of the square remains .
this means , that the spectrum of the quarter 2d sinai billiard is composed of two independent spectra , which differ by their parity with respect to the diagonal .
if we assume that the semiclassical deviations of the two spectra are not correlated , the above measure is the sum of the two independent measures .
it is plausible to assume also that both spectra have roughly the same semiclassical deviation , and thus @xmath193 is twice the semiclassical deviation of each of the spectra .
secondly , we recall that the 2d sinai billiard contains `` bouncing ball '' families of neutrally stable periodic orbits @xcite . we have subtracted their leading - order contribution from @xmath194 such that it includes ( to leading order ) only contributions from generic , isolated and unstable periodic orbits .
this is done since we would like to deduce from the 2d sinai billiard on the 2d generic case in which the bouncing balls are not present .
( in the sinai billiard , which is concave , there are also diffraction effects @xcite , but we did not treat them here . )
thirdly , the variant of ( [ eq : dtktau ] ) for billiards reads : @xmath195 when @xmath196 . in figure
[ fig : dikt.2 ] we demonstrate the compliance of the form factor with rmt goe using the integrated version of the above relation , and taking into account the presence of two independent spectra .
fourthly , it is interesting to know the actual numerical values of @xmath197 for the @xmath173 values that we considered .
we carried out the computation , and the results are presented in figure [ fig : ddelta2.2 ] .
it is interesting to observe , that for the entire range we have @xmath198 , which is very encouraging from an `` engineering '' point of view .
we now turn to the analysis of the 3d sinai billiard .
the billiard is the free space between a cube of edge @xmath164 and an inscribed sphere of radius @xmath165 , where @xmath199 ( see figure [ fig : sb2d - sb3d ] ) .
we used @xmath167 and @xmath200 and desymmetrized the billiard to the fundamental element ( 1/48 of the original one ) .
we calculated the lowest 6697 quantum levels in the interval @xmath201 and the shortest 586965 periodic orbits with length @xmath170 ( the number includes repetitions and time reversal conjugates ) . to treat the 3d sinai billiard we need to somewhat modify
the formalism which was presented in the previous section .
this is due to the fact that in 3d the contributions of the various non
generic bouncing
ball manifolds overwhelm the spectrum @xcite . since our goal is to give an indication of the semiclassical error in generic systems , it is imperative to get rid of this very strong non generic effect .
the bouncing ball amplitudes are @xmath202 where @xmath203 is the dimensionality of bouncing ball manifold in configuration space . in 3d @xmath204 typically , which completely overwhelms the contributions from isolated periodic orbits whose amplitude is @xmath205 .
even the diffraction corrections to the bouncing
ball amplitudes in 3d increase as @xmath206 with @xmath207 .
in contrast with the 2d case , however , it is difficult to subtract the bouncing ball contributions analytically for two main reasons .
first , there are always infinitely many bouncing ball primitive families in the 3d case , while in 2d there is only a finite number .
indeed , in any finite length interval there exists only a finite number of bouncing ball lengths , however their number for billiards in 3d exceeds by far the corresponding number for billiards in 2d .
second , the semiclassical amplitudes of the bouncing
balls are proportional to their volume in configuration space , and it is difficult in general to calculate it analytically .
thus , to overcome these difficulties we had to device a special method to cleanse the spectrum from the effect of the bouncing balls and from the leading diffractive corrections .
this method relies on the sensitivity of the eigenvalues to the boundary condition and it is described in detail in @xcite .
we shall briefly describe the essence of the method .
the most general ( mixed " ) boundary conditions under which the quantum billiard problem is self adjoint can be written as @xcite : @xmath208 where @xmath209 is the normal pointing outside of the billiard , the angle @xmath210 interpolates smoothly between dirichlet ( @xmath211 ) and neumann ( @xmath212 ) cases and @xmath213 has the dimension of a wavenumber .
note that @xmath210 and @xmath213 can be different for different parts of the boundary .
the spectrum depends parametrically on the boundary parameters , @xmath214 , and thus we can define the counting function @xmath215 . in @xcite
we discussed the 2d and 3d sinai billiards , and showed that if we choose dirichlet boundary conditions on all the billiard boundaries _ excluding _ the circle ( sphere ) , and set @xmath216 on the circle ( sphere ) then the quantity @xmath217 is to a large extent free of the effects of the bouncing balls . in the above @xmath218 which are the dirichlet everywhere eigenvalues .
the corresponding semiclassical trace formula reads @xcite : @xmath219 where @xmath220 in the above @xmath221 is the number of collisions with the circle ( sphere ) of the jth periodic orbit , and @xmath222 are the angles of incidence on the boundary with respect to the normal .
we note that ( [ eq : dqm - mbc ] ) is a weighted density of states where the standard @xmath223 weights of the @xmath224 functions are replaced by @xmath225 , and ( [ eq : derivmbc ] ) is a weighted trace formula where the standard amplitudes @xmath226 are replaced by @xmath227 .
one can show , that @xmath228 for long enough ( ergodic ) orbits .
we shall use @xmath229 for our purposes as follows . let us consider the weighted counting function : @xmath230 the function @xmath231 is a staircase with stairs of variable height @xmath225 . as was explained above , its advantage over @xmath18 is that it is semiclassically free of bouncing balls effects ( to leading order ) and corresponds only to the generic periodic orbits @xcite .
similarly , we construct from @xmath232 the function @xmath233 . having defined @xmath234
, we proceed in analogy to the dirichlet case .
we form from @xmath235 the functions @xmath236 , respectively , by multiplying with a window function @xmath237 and then construct the measure @xmath70 as in ( [ eq : delta2int ] ) .
the only difference is that the normalization of @xmath86 must be modified to account for the velocities " @xmath225 such as : @xmath238 the above considerations are meaningful provided that the `` velocities '' @xmath225 are narrowly distributed around a well - defined mean @xmath239 and we consider a small enough @xmath173-interval , such that @xmath239 does not change appreciably within this interval .
otherwise , @xmath70 is greatly affected by the fluctuations of @xmath225 ( which is undesired ) and the meaning of the normalization is questionable . to demonstrate the utility of the above construction using the mixed boundary conditions , we return to the 2d case .
we set @xmath240 , and note that the spectrum at our disposal for the mixed case was confined only to the interval @xmath241 .
first of all , we want to examine the width of the distribution of the @xmath225 s . in figure
[ fig : q ] we plot the ratio of the standard deviation of @xmath225 to the mean , averaged over the @xmath173-axis using a gaussian window .
we use the same window also in the calculations below .
the observation is that the distribution of @xmath225 is moderately narrow and the width decreases algebraically as @xmath173 increases .
this justifies the use of the mixed boundary conditions as was discussed above .
one also needs to check the validity of ( [ eq : dxkxi ] ) , and indeed we found compliance with goe also for the mixed case ( results not shown ) .
we next compare the functions @xmath242 for both the dirichlet and the mixed boundary conditions .
it turns out , that also in the mixed case the functions @xmath177 ( not shown ) fluctuate in @xmath179 with no special tendency .
the averages @xmath184 for the dirichlet and mixed cases are compared in figure [ fig : dmcavg.2 ] .
the values in the mixed case are systematically smaller than in the dirichlet case which is explained by the efficient filtering of tangent and close to tangent orbits that are vulnerable to large diffraction corrections @xcite . however , from @xmath243 on the two graphs show the same trends , and the values of @xmath244 in both cases are of the same magnitude .
thus , the qualitative behavior of @xmath113 is shown to be equivalent in the dirichlet and mixed cases , which gives us confidence in using @xmath245 together with the mixed boundary conditions procedure .
we finally applied the mixed boundary conditions procedure to compute @xmath89 for the desymmetrized 3d sinai with @xmath246 and set @xmath247 .
we first verified that also in the 3d case the velocities @xmath225 have narrow distribution see figure [ fig : q ] .
next , we examined equation ( [ eq : dxkxi ] ) using quantal data , and discovered that there are deviations form goe ( figure [ fig : mikt.3 ] ) .
we have yet no satisfactory explanation of these deviations , but we suspect that they are caused because the ergodic limit is not yet reached for the length regime under consideration due to the effects of the infinite horizon which are more acute in 3d .
nevertheless , from observing the figure as well as suggested by semiclassical arguments , it is plausible to assume that @xmath248 for small @xmath249 .
hence , this deviation should not have any qualitative effect on @xmath70 according to ( [ eq : delta2final ] ) .
similarly to the 2d case , the behavior of the function @xmath177 is fluctuative in @xmath179 , with no special tendency ( figure [ fig : mcxavg.3 ] ) .
if we average @xmath177 over the universal interval @xmath250 , we obtain @xmath184 which is shown in figure [ fig : mcavg.3 ] .
the averages @xmath184 are fluctuating with a mild decease in @xmath173 , and therefore we can conclude that @xmath251 where the `` optimistic '' measure ( leftmost term ) corresponds to @xmath252 , and the `` pessimistic '' one ( rightmost term ) is due to @xmath253 . in other words , the error estimates ( [ eq : delta2 - 2d ] , [ eq : delta2 - 3d ] ) for the 2d and the 3d cases , respectively , are the same , and in sharp contrast to the `` traditional '' error estimate which predicts that the errors should be different by a factor @xmath254 . on the basis of our numerical data , and in spite of the uncertainties which were clearly delineated , we can safely exclude the traditional " error estimate .
our main finding was that the upper bound on the semiclassical error is a logarithmic divergence , which is independent of the dimension ( equations ( [ eq : delta2 - 2d ] ) , ( [ eq : delta2 - 3d ] ) ) . in this respect , there are a few points which deserve discussion . to begin , we shall try to evaluate @xmath89 using the explicit expressions for the leading corrections to the semiclassical counting function of 2d generic billiard system , as derived by alonso and gaspard @xcite : @xmath255\ ] ] where @xmath226 are the standard semiclassical amplitudes ( see ( [ eq : gtf ] ) ) , @xmath256 are the lengths of periodic orbits and @xmath257 are the @xmath173-independent amplitudes of the @xmath258 corrections .
the @xmath257 s are given in @xcite .
we ignored in the above equation the case of odd maslov indices .
if we calculate from @xmath259 the corresponding length spectrum @xmath260 using a ( normalized ) gaussian window @xmath261{\pi \sigma^2 } ) \exp [ -(k'-k)^2/(2 \sigma^2 ) ] $ ] , we obtain : @xmath262{\pi } } \sum_j \frac{a_j}{l_j } \left [ { \rm e}^{i k ( l - l_j ) - i \frac{q_j}{k } } { \rm e}^{-(l - l_j)^2 \frac{\sigma^2}{2 } } - { \rm e}^{i k ( l + l_j ) + i \frac{q_j}{k } } { \rm e}^{-(l + l_j)^2 \frac{\sigma^2}{2 } } \right ] .\ ] ] in the above we regarded the phase @xmath263 as slowly varying .
the results of alonso and gaspard @xcite suggest that the @xmath257 are approximately proportional to the length of the corresponding periodic orbits : @xmath264 we can therefore well
approximate @xmath194 as : @xmath265{\pi } } { \rm e}^{-i q
l / k } \sum_j \frac{a_j}{l_j } \left [ \cdots \right ] = { \rm e}^{- i q l / k } \hat{d}_{\rm sc - gtf } \
, , \ ] ] where @xmath266 is the length spectrum which corresponds to the semiclassical gutzwiller trace formula for the counting function ( without @xmath258 corrections ) .
we are now in a position to evaluate the semiclassical error , indeed : @xmath267 if we now use equation ( [ eq : dxkxi ] ) and @xmath268 ( which is valid for @xmath269 for chaotic systems ) , we get : @xmath270 for @xmath159 we have that @xmath271 which is negligible , hence we can replace the lower limit in ( [ eq : delta2-analytical1 ] ) with 0 : @xmath272 this is the required expression .
the dimensionality enter in @xmath273 only through the power of @xmath173 in @xmath274 .
let us apply equation ( [ eq : delta2-analytical2 ] ) to 2d and 3d cases .
for 2d we have that @xmath275 in leading order , where @xmath163 is the billiard s area , thus , @xmath276 which means , that the semiclassical error in 2d billiards is of the order of the mean spacing , and therefore the semiclassical trace formula is ( marginally ) accurate and meaningful .
this is compatible with our numerical findings within the limitations of the numerical fluctuations . for 3d , the coefficients @xmath257
were not obtained explicitly , but we shall assume that they are still proportional to @xmath256 ( equation ( [ eq : mj - lj ] ) ) and thus that ( [ eq : delta2-analytical2 ] ) holds . for 3d billiards @xmath277 to leading order , where @xmath278 is the billiard s volume .
thus the upper limit in ( [ eq : delta2-analytical2 ] ) is @xmath279 which is large in the semiclassical limit . in that case
, we can replace @xmath280 with its mean value @xmath281 and the integrand becomes essentially @xmath282 which results in : @xmath283 that is , in contrast to the 2d case , the semiclassical error diverges logarithmically and the semiclassical trace formula becomes meaningless as far as the prediction of individual levels is concerned .
this is compatible with our numerical results within the numerical dispersion .
however , it relies heavily on the assumption that @xmath284 , for which we can offer no justification .
we note in passing , that the logarithmic divergence persists also for @xmath285 .
another interesting point relates to integrable systems .
it can happen , that for an integrable system it is either difficult or impossible to express the hamiltonian as an explicit function of the action variables . in that case
, we can not assign to the levels other quantum numbers than their ordinal number , and the semiclassical error can be estimated using @xmath70 . however , since for integrable systems @xmath286 , we get that : @xmath287 therefore , for deviations which are comparable to the chaotic cases , @xmath288 , we get @xmath289 which is much larger than for the chaotic case and diverges for @xmath290 .
the formula ( [ eq : delta2final ] ) for the semiclassical error contains semiclassical information in two respects . obviously , @xmath148 , which is the difference between the quantal and the semiclassical length spectra contains semiclassical information .
but also , the fact that the lower limit of the integral in ( [ eq : delta2final ] ) is finite is a consequence of semiclassical analysis . if this lower limit is replaced by @xmath73 , the integral diverges for finite values of @xmath11 , which is meaningless
therefore , the fact that the integral has a lower cutoff , or rather , that @xmath291 is exactly @xmath73 below the shortest period , is a crucial semiclassical ingredient in our analysis .
finally , we consider the case in which the semiclassical error is estimated with no periodic orbits taken into account .
that is , we want to calculate @xmath292 which is the number variance @xmath293 for the large argument @xmath294 .
this implies @xmath295 , and using ( [ eq : delta2final ] ) we get that @xmath296 , which in the semiclassical limit becomes @xmath297 .
this result is fully consistent and compatible with previous results for the asymptotic ( saturation ) value of the number variance @xmath298 ( see for instance @xcite ) .
it implies also that the pessimistic error bound ( [ eq : delta2-pessimistic ] ) is of the same magnitude as if periodic orbits were not taken into account at all .
( periodic orbits improve , however , quantitatively , since in all cases we obtained @xmath299 . )
thus , if we assume that periodic orbit contributions do not make @xmath300 worse than @xmath301 , then the pessimistic error bound @xmath302 is the _ maximal _ one in any dimension @xmath2 .
this excludes , in particular , algebraic semiclassical errors , and thus refutes the traditional estimate @xmath303 .
this work was finished when hp held a minerva post doctoral fellowship in freiburg university and us spent a sabbatical leave at the isaac newton institute , cambridge .
both authors thank their respective hosts for the hospitality and support .
we would like to acknowledge discussions with and comments from michael berry , reinhold blmel , doron cohen , predrag cvitanovic , bruno eckhardt , martin gutzwiller and frank steiner .
the work was supported by the minerva center for physics of complex systems and by grants from the israel science foundation .
10 martin c. gutzwiller . the semi - classical quantization of chaotic hamiltonian systems . in m .-
giannoni , a. voros , and j. zinn - justin , editors , _ proceedings of the 1989 les houches summer school on `` chaos and quantum physics '' _ , pages 201249 .
elsevier science publishers b.v . ,
amsterdam , 1991 .
p. m. bleher .
quasi - classical expansions and the problem of quantum chaos . in _ geometric aspects of functional analysis ,
semin . , gafa , isr .
1989 - 90 , lect .
notes math .
_ , pages 6089 .
springer , 1991 .
j. p. keating . the riemann zeta function and quantum chaology . in g. casati , i. guarnery , and u. smilansky , editors ,
_ proceedings of the 1991 enrico fermi international school on `` quantum chaos '' , course cxix_. north - holland , 1993 . m. v. berry .
some quantum to classical asymptotics . in m .- j .
giannoni , a. voros , and j. zinn - justin , editors , _ proceedings of the 1989 les houches summer school on `` chaos and quantum physics '' _ , page 251 .
elsevier science publishers b.v . , amsterdam , 1991 . | the semiclassical trace formula provides the basic construction from which one derives the semiclassical approximation for the spectrum of quantum systems which are chaotic in the classical limit .
when the dimensionality of the system increases , the mean level spacing decreases as @xmath0 , while the semiclassical approximation is commonly believed to provide an accuracy of order @xmath1 , _ independently _ of @xmath2 .
if this were true , the semiclassical trace formula would be limited to systems in @xmath3 only . in the present work we set about to define proper measures of the semiclassical spectral accuracy , and to propose theoretical and numerical evidence to the effect that the semiclassical accuracy , measured in units of the mean level spacing , depends only weakly ( if at all ) on the dimensionality .
detailed and thorough numerical tests were performed for the sinai billiard in @xmath4 and @xmath5 dimensions , substantiating the theoretical arguments . |
massive stars play a key role in our universe .
they drive the chemical evolution of galaxies by synthesising most of the heavy elements . their strong stellar winds , radiation feedback , powerful supernova explosions and long gamma ray bursts shape the interstellar medium .
they are thought to have played an essential role in reionising the universe after the dark ages and are visible up to large distances .
unfortunately , our understanding of the formation and evolution of the most massive stars in the local universe is incomplete .
recently it was established that most of the massive stars in the milky way are actually part of a binary star system and that more than @xmath0 of them will exchange mass with a companion during their life @xcite .
our understanding of these stars is further hampered by two major controversies .
the first one , the cluster age problem , concerns the ages of the youngest star clusters . emerging star clusters
are expected to form stars in a time span shorter than the lifetime of their most massive members @xcite .
in contrast , the most luminous stars in two of the richest young clusters in our galaxy , the arches and quintuplet clusters , show an apparently large age range .
their hydrogen- and nitrogen - rich wolf - rayet ( wnh ) stars appear significantly younger than most of their less luminous o stars .
similar age discrepancies are observed in other young stellar systems such as the cygnus ob2 association and the star clusters pismis 24 and ngc 6611 . the second controversy , the maximum stellar mass problem , concerns the stellar upper mass limit . such a upper mass limit is theoretically motivated by the eddington - limit which may prevent stellar mass growth by accretion above a certain mass .
observationally , an upper mass limit of about @xmath7 is derived from the individual stellar mass distributions of the arches and the r136 clusters @xcite and from a broader analysis of young stellar clusters @xcite .
this result is questioned by a recent analysis of very massive stars in the core of r136 , in which stars with initial masses of up to about @xmath9 are found @xcite .
furthermore , recently detected ultra - luminous supernovae in the local universe are interpreted as explosions of very massive stars e.g. sn 2007bi is well explained by a pair - instability supernova from an initially @xmath10 star @xcite . here
, we show that both controversies can be resolved by considering a time dependent stellar mass function in young star clusters that accounts for stellar wind mass loss and binary mass exchange .
we perform detailed population synthesis calculations of massive single and binary stars that include all relevant physical processes affecting the stellar masses and compare them to observed present day mass functions of the arches and quintuplet clusters .
our methods and the observations of the mass functions of the arches and quintuplet clusters are described in sec .
[ sec : methods ] .
we analyse the arches and quintuplet clusters in sec .
[ sec : analysis - arches - quintuplet ] to derive cluster ages and identify possible binary products by fitting our models to the observed mass functions .
stochastic sampling effects are investigated in sec .
[ sec : stochastic - sampling - effects ] and the implications of our findings for the upper mass limit are explored in sec .
[ sec : upper - mass - limit ] .
we discuss our results in sec .
[ sec : discussion ] and give final conclusions in sec .
[ sec : conclusions ] .
we analyse the arches and quintuplet clusters in two steps .
first , we model their observed stellar mass functions to e.g. determine the initial mass function ( imf ) slopes and the cluster ages .
we set up a dense grid of single and binary stars , assign each stellar system in the grid a probability of existence given the initial distribution functions ( cf .
[ sec : init - distr - fcts ] ) and evolve the stars in time using our rapid binary evolution code described ( cf .
[ sec : binaryc ] ) .
present - day mass functions are then constructed from the individual stellar masses at predefined ages .
this ensures that all the relevant physics like stellar wind mass loss and binary mass exchange , which directly affects stellar masses , is factored in our mass functions .
second , we investigate stochastic sampling effects to e.g. compute the probability that the most massive stars in the arches and quintuplet clusters are binary products . to that end , we randomly draw single and binary stars from initial distribution functions until the initial cluster masses are reached and , again , evolve the drawn stellar systems with our rapid binary evolution code . the set - up of these monte carlo experiments
is described in detail in sec .
[ sec : mc - experiments ] .
the initial distribution functions used in the above mentioned steps are summarised in sec .
[ sec : init - distr - fcts ] and an overview of the observations of the arches and quintuplet clusters to which we compare our models is given in sec . [
sec : observations ] .
we bin mass functions in a non - standard way to compare them to observations the binning procedure is described in sec .
[ sec : binning - procedure ] .
the details of our population synthesis code are described in @xcite and @xcite . here , we briefly summarise the most important methods and assumptions that are used to derive our results .
we use a binary population code to evolve single and binary stars and follow the evolution of the stellar masses and of other stellar properties as a function of time .
our code is based on a rapid binary evolution code @xcite which uses analytic functions @xcite fitted to stellar evolutionary models with convective core overshooting @xcite to model the evolution of single stars across the whole hertzsprung - russell diagram .
we use a metallicity of @xmath13 .
stellar wind mass loss is applied to all stars with luminosities larger than @xmath14 @xcite .
the mass accretion rate during mass transfer is limited to the thermal rate of the accreting star .
binaries enter a contact phase and merge if the mass ratio of accretor to donor is smaller than a critical value at the onset of roche lobe overflow @xcite .
when two main sequence stars merge , we assume that @xmath15 of the total mass is lost and that @xmath15 of the envelope mass is mixed with the convective core @xcite . photometric observations of star clusters can not resolve individual binary components . in order to compare our models to observations
we assume that binaries are unresolved in our models and determine masses from the combined luminosity of both binary components utilising our mass - luminosity relation .
hence unresolved , pre - interaction binaries contribute to our mass functions .
we concentrate on main sequence ( ms ) stars because stars typically spend about @xmath16 of their lifetime in this evolutionary phase ; moreover , our sample stars used for comparison are observationally colour - selected to remove post - ms objects .
if a binary is composed of a post - ms and a ms star , we take only the ms component into account .
we assume that primary stars in binaries and single stars have masses @xmath17 distributed according to a power law initial mass function ( imf ) with slope @xmath18 , @xmath19 in the mass range @xmath20 to @xmath21 ( where @xmath22 is a normalisation constant ) .
secondary star masses , @xmath23 , are taken from a flat mass ratio distribution , i.e. all mass ratios @xmath24 are equally probable @xcite .
the initial orbital periods for binaries with at least one o - star , i.e. a primary star with @xmath25 , mass ratio @xmath26 and a period @xmath27 are taken from the distribution of stars in galactic open clusters @xcite .
the initial periods of all other binaries follow a flat distribution in the logarithm of the orbital period @xcite .
orbital periods are chosen such that all interacting binaries are taken into account , i.e. the maximum initial orbital separation is @xmath28 ( @xmath29 ) .
binaries with wider orbits would be effectively single stars . to address stochastic sampling , we perform monte carlo simulations of star clusters and investigate the probability that the most massive star in a star cluster is a product of binary evolution as a function of cluster mass , age , binary fraction and imf slope .
we assume that all stars are coeval and that every star cluster forms from a finite supply of mass with stellar masses stochastically sampled from initial distribution functions .
while single stars are sampled from the initial mass function , binary stars are chosen from a larger parameter space of primary and secondary masses and orbital periods .
this larger parameter space is better sampled in clusters of higher mass .
we draw single and binary stars for a given binary fraction from the initial distribution functions of primary mass , secondary mass and orbital period until a given initial cluster mass , @xmath30 , is reached . here
we consider only stars with masses in the range of @xmath20@xmath21 .
the true cluster masses are therefore larger if stars below @xmath31 are added . including these stars according to a kroupa imf @xcite , increases the true cluster mass by @xmath32 and @xmath33 for high mass ( @xmath34 ) imf slopes of @xmath35 and @xmath36 , respectively .
after the stellar content of a cluster is drawn , we evolve the stars in time to analyse whether the most massive stars at a given cluster age result from close binary interaction . repeating this experiment @xmath37 times
provides the probability that the most massive cluster star formed from binary interactions , how long it takes on average until the most massive star is a product of binary evolution , @xmath38 , and how many stars have on average a mass larger than that of the most massive cluster star ( @xmath39 ) which did not accrete from a companion , @xmath40 .
the most massive star that did not accrete from a companion can be a genuine single star or a star in a binary where binary mass transfer has not yet happened . from here on we refer to this star as ` the most massive genuine single star ' .
we evolve and distribute stars as described in secs .
[ sec : binaryc ] and [ sec : init - distr - fcts ] . to compare our monte carlo simulations with observations of the arches and quintuplet clusters , we need to know the corresponding cluster masses @xmath41 in our monte carlo experiments .
we use imf slopes of @xmath42 as later determined in sec .
[ sec : comparison - mass - functions ] for both clusters .
the observations used for comparison are complete for masses @xmath43 , corresponding to @xmath44 and @xmath45 stellar systems in the arches and quintuplet and integrated masses of stars more massive than @xmath46 of @xmath47 and @xmath48 , respectively .
our best fitting monte carlo models of the central regions of arches and quintuplet with primordial binary fractions of @xmath49 and @xmath50 have @xmath51 stellar systems with an integrated ( initial ) mass of @xmath52 and @xmath53 stellar systems with an integrated initial mass of @xmath54 , respectively .
these models correspond to initial cluster masses of @xmath55 and @xmath56 , respectively , in stars with @xmath57 .
we assume that binaries are resolved in our monte carlo calculations , contrary to when we model mass functions in order to compare to observed mass functions .
this is because we make theoretical predictions and are thus interested in individual masses of all stars regardless of them being in a binary or not .
the observed present - day mass functions of the arches and quintuplet clusters were obtained from naos / conica ( naco ) photometry at the vlt . the arches cluster was observed in 2002 over a field of view ( fov ) of @xmath58 .
the centre of the quintuplet cluster was imaged with naco over a fov of @xmath59 in 2003 and 2008 , which allowed the construction of a membership source list from proper motions .
both data sets were obtained in the @xmath60 ( @xmath61 ) and @xmath62 ( @xmath63 ) passbands .
the colour information is used to remove likely blue foreground interlopers , red clump and giant stars towards the galactic center line of sight .
details can be found in @xcite for the arches cluster and in for the quintuplet cluster . in the case of the arches cluster ,
the known radial variation of the extinction is removed prior to individual mass determination employing the extinction law of @xcite .
masses are then derived from the @xmath62 magnitudes of each star by comparison with a @xmath64 geneva isochrone . in the case of the quintuplet , the better photometric performance allowed all sources to be individually dereddened to a @xmath65 padova ms isochrone using the recently updated near - infrared extinction law towards the galactic center line of sight @xcite .
as detailed in , isochrone ages of @xmath66 and @xmath67 do not significantly alter the shape and slope of the constructed mass function .
all mass determinations are based on solar metallicity evolution models . with the aim to minimise any residual field contamination , only the central @xmath68 or @xmath69 of the arches and @xmath70 or @xmath71 of the quintuplet ( at an assumed distance of @xmath72 to the galactic center ; @xcite ) were selected to construct the mass functions .
for the arches cluster , this radial selection corresponds approximately to the half - mass radius , which implies that the mass projected into this annulus is of the order of @xmath73 . in the quintuplet cluster
, the total mass is estimated to be @xmath74 within the considered @xmath71 radius .
the mass functions in the central regions of arches and quintuplet have slopes that are flatter than the usual salpeter slopes , most likely because of mass segregation .
the most massive stars in the arches and quintuplet clusters are hydrogen and nitrogen rich wnh stars . as reliable masses
can not be derived for these wolf - rayet ( wr ) stars from photometry alone , and as several of the wrs in the quintuplet suffered from saturation effects , the most massive stars are excluded from the mass functions .
this affects @xmath75 wnhs stars with uncertain masses in arches and @xmath66 ( plus @xmath76 post - ms , carbon rich wr stars ) in quintuplet .
these wnh stars are expected to contribute to the high mass tail of the arches and quintuplet mass functions .
following @xcite , we employ a binning procedure that renders the observed mass functions independent of the starting point of the bins .
we shift the starting point by one tenth of the binsize and create mass functions for each of these starting points .
we use a fixed binsize of @xmath77 to ensure that the number of stars in each bin is not too small and does not introduce a fitting bias @xcite .
each of these ten mass functions with different starting points is shown when we compare our mass functions to observations .
this procedure results in lowered number counts in the highest mass bins because only the most massive stars will fall into these bins as seen in the power - law mass function ( black dotted lines in fig .
[ fig : observations ] ) where a kink is visible around @xmath78 ( left panel ) and @xmath79 ( right panel ) respectively ( cf . convolution of a truncated horizontal line with a box function with the width of the bins ) .
this kink is caused by the binning procedure .
importantly , the observations , our models and the power - law mass functions in figs .
[ fig : observations ] and [ fig : sfh ] are binned identically to render the mass functions comparable .
for a meaningful comparison of the modelled with the observed mass functions , the star cluster and the observations thereof need to fulfil certain criteria . they should be * between @xmath80 and @xmath81 in age such that the wind mass loss peak in the mass function is present ( see below and * ? ? ?
* ) , * massive enough such that the mass function samples the largest masses * homogeneously analysed , with a complete _ present - day _ mass function above @xmath82 .
both , the arches and quintuplet clusters fulfil all criteria and are therefore chosen for our analysis . other possible star clusters , which can be analysed in principle , are the galactic center cluster , ngc 3603 yc , westerlund 1 and r136 in the large magellanic cloud .
trumpler 14 and trumpler 16 in the galactic carina nebula are not massive enough and rather an ob star association with stars of different ages , respectively . for westerlund 1 @xcite and ngc 3603 yc @xcite ,
present - day mass functions were recently derived .
a brief inspection of these results shows that both clusters may be suitable for an analysis as performed here for the arches and quintuplet clusters .
we will investigate this further in the near future .
possibly , ngc 3603 yc is too young such that its mass function is not yet altered enough by stellar evolution to apply our analysis .
the initially most massive stars in a cluster end their life first .
this depopulates the high mass end of the stellar mass function . before that , however , massive stars lose a significant fraction of their initial mass because of strong stellar winds ; e.g. our @xmath21 star at solar metallicity loses about @xmath83 during core hydrogen burning .
stellar wind mass loss shifts the top of the mass function towards lower masses and a peak accumulates near its high mass end ( figs .
[ fig : models]a and [ fig : models]b ) .
the location of the peak depends strongly on the cluster age and provides a clock to age - date a star cluster .
stars in close binary systems exchange mass with their companion either by mass transfer or in a stellar merger . _
a fraction _ of stars gain mass , producing a tail at the high mass end of the mass function ( hatched regions in fig .
[ fig : models]b ) which extends beyond the most massive single - stars ( fig .
[ fig : models]a ) by up to a factor of about two . the mass gainers _ appear _ younger than genuine single stars because their convectively mixed stellar core grows upon mass accretion and mixes fresh fuel into their centre , thereby turning their clock backwards .
furthermore , the most massive gainers reach masses which , when interpreted as single stars , have lifetimes that are shorter than the cluster age they are the massive counterpart of classical blue straggler stars @xcite .
the mass functions of the cores of the arches ( @xmath84 ) and quintuplet ( @xmath85 ) clusters reveal both the stellar wind mass loss peak and the tail because of binary mass exchange . compared to a power - law , we find that the arches and quintuplet mass functions are overpopulated in the ranges @xmath86 to @xmath87 ( @xmath88@xmath89 ) and @xmath90 to @xmath91 ( @xmath92@xmath93 ) , respectively ( fig . [
fig : observations ] ) .
these peaks are well reproduced by our models ( figs .
[ fig : models ] and [ fig : observations ] ) .
we can thus determine the cluster age because among the stars in the peak are the initially most massive stars that are ( a ) still on but about to leave the main sequence and ( b ) unaffected by binary interactions ( we refer to them as turn - off stars in analogy to their position close to the turn - off in a hertzsprung - russell diagram ) .
the majority of stars in the peak are turn - off stars but there are small contributions from unresolved and post - interaction binaries ( see * ? ? ?
a correction for wind mass loss then reveals the initial mass of the turn - off stars and hence the age of the cluster .
we can correct for wind mass loss by redistributing the number of excess stars in the peak @xmath94 such that the mass function is homogeneously filled for masses larger than those of the peak stars up to a maximum mass , the initial mass of the turn - off stars @xmath95 .
the number of excess stars is then @xmath96 where @xmath97 is the present - day mass of the turn - off stars , which can be directly read - off from the upper mass end of the peak , and @xmath98 the initial mass function as defined in eq .
( [ eq : gammadef ] ) .
the initial mass of the turn - off stars @xmath95 and hence the cluster age follows from integrating eq .
( [ eq : number - excess - stars ] ) , @xmath99 the normalisations , @xmath22 , of the mass functions to be filled up with the excess stars , @xmath94 , ( dotted , power - law functions in fig . [ fig : observations ] ) are @xmath100 and @xmath101 for arches and quintuplet , respectively , with slopes of @xmath42 in both cases ( see discussion below for why the mass functions are so flat ) .
it is difficult to read - off the exact value of @xmath97 from the observed mass functions because of the binning .
but from binning our modelled mass functions in the same way as the observations , we know that @xmath97 corresponds to the mass shortly after the peak reached its local maximum ( cf . the vertical dashed lines in figs . [ fig : models ] and [ fig : observations ] ) .
depending on the exact value of @xmath97 , @xmath102@xmath103 in arches and @xmath104@xmath105 in quintuplet , we find @xmath106@xmath107 and @xmath76@xmath108 excess stars in the peaks of the arches and quintuplet mass functions , respectively .
these numbers of excess stars result in turn - off masses @xmath95 of @xmath109@xmath110 and @xmath111@xmath112 and hence ages of @xmath113@xmath114 and @xmath115@xmath116 for the arches and quintuplet clusters , respectively .
these are only first , rough age estimates that will be refined below and their ranges stem from the uncertainty in reading - off @xmath97 from the observed mass functions . from the difference between the initial and present day masses of the turn - off stars in arches and quintuplet
, we can directly measure the amount of mass lost by these stars through stellar winds .
the turn - off stars in arches lost about @xmath106@xmath117 and the turn - off stars in quintuplet about @xmath118@xmath119 during their ms evolution .
this is a new method to measure stellar wind mass loss which does not require measurements of stellar wind mass loss rates and can therefore be used to constrain these .
more accurately , we determine the ages of the arches and quintuplet clusters by fitting our population synthesis models ( sec . [
sec : methods ] ) to the observed mass functions .
first , we fit power - law functions to the observed mass functions in mass regimes in which they are observationally complete and not influenced by stellar wind mass loss ( @xmath120 and @xmath121 , respectively ) .
binary effects are also negligible because stars with such masses are essentially unevolved at the present cluster ages .
this fit gives the normalisation and a first estimate of the slope of the mass function .
we then vary the mass function slope , the cluster age and the primordial binary fraction in our models simultaneously such that the least - square deviation from the observations is minimised .
our best - fit models are shown in fig .
[ fig : observations ] together with the observed mass functions .
we find slopes of @xmath42 , ages of @xmath122 and @xmath123 and primordial binary fractions of @xmath49 and @xmath50 for the arches and quintuplet cluster , respectively .
the binary fractions are less robust and may be the same within uncertainties because we do not take the uncertain masses of the wnh stars into account . while our mass function fits contribute to the age uncertainties by only @xmath124 , its major part , @xmath125 and @xmath126 for arches and quintuplet , respectively , is due to observational uncertainties in stellar masses of @xmath127 ( sec .
[ sec : discussion ] ) .
massive stars tend to sink towards the cluster cores because of dynamical friction ( mass segregation ) , thereby flattening the mass function of stars in the core .
the derived mass function slopes of @xmath42 are flatter than the typical salpeter slope of @xmath36 @xcite because we investigate only the mass segregated central regions of both clusters , i.e. a subsample of stars biased towards larger masses . in our models ( fig . [ fig : models ] ) , the tail of the arches mass function contains about @xmath128 unresolved , pre - interaction binaries with @xmath129 ( @xmath130 ) and about @xmath32 with @xmath131 ( @xmath132 ) . for quintuplet ,
the fraction of unresolved , pre - interaction binaries is about @xmath32 with @xmath133 ( @xmath134 ) and about @xmath15 with @xmath135 ( @xmath136 ) .
the binary fraction among the rejuvenated binary products in the tails is about @xmath137 in our arches and @xmath0 in our quintuplet model , where the remaining stars are single star binary products , i.e. merger stars .
previously estimated ages for the arches and quintuplet clusters lie in the range @xmath138@xmath139 and @xmath138@xmath140 , respectively . within these ranges ,
the age discrepancy between the most luminous cluster members , the wn and the less luminous o stars , accounts for about @xmath141 and @xmath142 , respectively , which is eliminated by our method .
our ages of @xmath122 and @xmath123 for the arches and quintuplet clusters , respectively , agree with the ages derived from the o stars and dismiss the proposed younger ages from the brightest stars as a result of neglecting binary interactions .
the most famous member of the quintuplet , the pistol star , is such an example because it appears to be younger than @xmath143 assuming single - star evolution @xcite .
the initial mass of the primary star , the mass ratio and the orbital period of a binary system determine when mass transfer starts , with more massive and/or closer binaries interacting earlier .
stochastic effects caused by the limited stellar mass budget prevent the formation of all possible binaries in a stellar cluster , i.e. binaries with all possible combinations of primary mass , mass ratio and orbital period .
the likelihood that a binary in a given cluster interacts , e.g. after @xmath64 , and that the binary product becomes then the most massive star depends thus on the number of binary stars in that cluster , hence on the total cluster mass . using monte carlo simulations , we investigate the influence of stochastic sampling and binary evolution on the most massive stars in young star clusters ( cf .
[ sec : mc - experiments ] ) . the galactic star cluster ngc 3603yc contains ngc 3603-a1 , a binary star with component masses @xmath144 and @xmath145 in a @xmath146 orbit @xcite .
an initially @xmath147 binary in a @xmath146 orbit starts mass transfer @xmath148 after its birth according to the non - rotating models of .
this is the time needed for the @xmath149 primary star to fill its roche lobe as a result of stellar evolutionary expansion .
this time provides an upper age estimate for ngc 3603 yc . after mass transfer
, the secondary star will be the most massive star in the cluster .
were ngc 3603-a1 in a closer orbit , it could already be a binary product today . to find the probability that the most massive star in a cluster of a given age is a binary product , we investigate how many close binaries are massive enough to become the most massive star by mass transfer .
were the cluster a perfect representation of the initial stellar distribution functions , we could use these functions to derive the probability directly .
however , the finite cluster mass and hence sampling density must be considered for comparison with real clusters .
returning to the example of ngc 3603 yc , were the cluster of larger total mass , its binary parameter space would be better sampled and its most massive star might already be a binary interaction product . for perfect sampling , i.e. infinite cluster mass , the time until a binary product is the most massive star tends towards zero .
the idea that the most massive star in a star cluster may be a binary product resulted from the first discovery of blue straggler stars @xcite .
it was proposed that blue stragglers might stem from binary mass transfer and/or stellar collisions @xcite .
stellar population synthesis computations including binary stars then showed that this is indeed possible . here , we show using the binary distribution functions of @xcite that the formation of blue stragglers by binary interactions prevails up to the youngest and most massive clusters and quantify it for the arches and quintuplet clusters . in fig .
[ fig : average - times - until - interaction ] , we show the average time @xmath38 after which the most massive star in a star cluster is a product of binary evolution as a function of the cluster mass @xmath30 for two different primordial binary fractions @xmath150 .
the error bars are @xmath151 standard deviations of @xmath37 monte carlo realisations .
the slope of the mass function is @xmath42 , appropriate for the mass - segregated central regions of both arches and quintuplet . the more massive a star cluster ,
i.e. the more stars populate the multidimensional binary parameter space , the shorter is this average time because the probability for systems which interact early in their evolution is increased . for less massive clusters @xmath38 increases and the statistical uncertainty grows .
for example if @xmath152 , there are only about @xmath153 binaries in which at least one star has a mass above @xmath46 ( for @xmath42 and @xmath154 ) .
the same reasoning holds for different binary fractions : the higher the binary fraction , the more binaries and hence a shorter average time until the most massive star results from binary interactions . until the most massive star in a star cluster is a product of close binary evolution as a function of cluster mass for two primordial binary fractions , @xmath155 , and a mass function slope @xmath42 . for steeper ,
salpeter - like mass functions see fig .
[ fig : average - times - until - interaction-2 ] .
the error bars are the standard deviation of @xmath37 realisations of each cluster .
the star symbols indicate the age and _ central _ cluster mass of arches and quintuplet as derived in this work.,scaledwidth=46.0% ] with a salpeter mass function ( @xmath36 , * ? ? ?
* ) the average time until the most massive star is a binary product increases compared to @xmath42 ( fig .
[ fig : average - times - until - interaction-2 ] ) because there are fewer massive binaries that interact to form the most massive star .
assume a @xmath65 old star cluster has a mass function slope of @xmath36 , a total mass in stars above @xmath31 of @xmath156 ( i.e. a true cluster mass of @xmath157 if stars below @xmath31 follow a kroupa imf ; see sec . [ sec : mc - experiments ] ) and a primordial binary fraction of @xmath158 . from fig .
[ fig : average - times - until - interaction-2 ] , we can then read - off after which time the most massive star is expected to be a binary product , namely after @xmath159 .
but for a steeper mass function with a salpeter slope of @xmath36 .
the binary parameter space spanned by the initial mass ratios and initial orbital separations for massive primary stars is now less populated , resulting in increased average times until the most massive star is a binary product .
similarly , the standard deviations increase.,scaledwidth=46.0% ] the central regions of the arches and quintuplet clusters have masses of @xmath160 and @xmath161 in stars more massive than @xmath31 ( sec .
[ sec : mc - experiments ] ) and ages of @xmath162 and @xmath123 ( sec .
[ sec : comparison - mass - functions ] ) , respectively . from fig .
[ fig : average - times - until - interaction ] , we expect that the most massive star in the arches cluster is a binary product after @xmath163 and after @xmath164 in the quintuplet cluster . in fig .
[ fig : prob - numbers ] we show the probability that the most massive star is a binary product and the average number of stars that are more massive than the most massive genuine single star as a function of cluster age for different cluster masses and two different binary fractions .
the imf slope is @xmath42 .
the corresponding probabilities and average numbers for a salpeter ( @xmath36 ) mass function are shown in fig .
[ fig : prob - numbers-2 ] .
again , the error bars are @xmath151 standard deviations from @xmath37 monte carlo experiments . returning to the above mentioned example star cluster ( @xmath156 ) : from fig .
[ fig : prob - numbers-2 ] , we find that the most massive star is a binary product with a probability of @xmath165 and that the most massive @xmath166 stars are expected to be binary products for the exemplary cluster age of @xmath65 . given the ages of the arches and quintuplet clusters , we find a probability of @xmath167 that the most massive star in each cluster is a binary product , with the most massive @xmath168 and @xmath169 stars being products of binary evolution in arches and quintuplet , respectively .
this is compatible with the number of wnh stars in the cores of arches and quintuplet , which are the most luminous and hence most massive stars in these clusters , implying they are massive blue stragglers .
data from two star clusters provide the current evidence for the existence of an upper stellar mass limit around @xmath7 : the arches cluster in the galactic center @xcite and the r136 cluster in the large magellanic cloud @xcite .
however , according to our analysis an upper mass limit can not be derived from the arches cluster because ( a ) it is too old , hence the most massive stars already exploded , and ( b ) its present - day high mass star population is dominated by binary products .
the situation might be different in the r136 cluster : current age estimates lie in the range @xmath20@xmath170 @xcite . in the following
we assume that the cluster is young enough such that even the most massive stars have not yet evolved off the main sequence , to explore what we can learn from r136 about a possible stellar upper mass limit .
four stars in r136 with initial masses of @xmath8@xmath9 appear to exceed the currently discussed upper mass limit of @xmath7 @xcite . either these stars were born with masses exceeding @xmath7 or gained mass from other stars e.g. by binary interactions ( this work ) or dynamically induced stellar mergers . from our monte carlo simulations ( sec .
[ sec : stochastic - sampling - effects ] ) , we can not judge with high enough confidence whether the most massive star in r136 is expected to be a binary product or not because of the uncertain age of r136 .
r136 has an imf with approximately a salpeter slope @xmath36 @xcite and its cluster mass is @xmath171@xmath172 . from our monte carlo simulations of star clusters with binary fractions of @xmath50 and cluster masses
@xmath173 of @xmath174 and @xmath175 ( fig .
[ fig : prob - numbers-2 ] ) , we find that the most massive star is expected to be a binary product after @xmath141 with a probability of @xmath176 and @xmath177 , respectively .
the probabilities increase to @xmath178 and @xmath179 , respectively , for a cluster age of @xmath64 and are larger than @xmath180 for an age of @xmath181 .
so if the cluster is older than about @xmath64 , the most massive star is likely a binary product ( note that our calculations are for a metallicity of @xmath13 while the r136 cluster in the large magellanic cloud has a lower metallicity so the above numbers will slightly change for the appropriate metallicity but are good enough for this estimate ) .
because it is not clear whether the most massive star in r136 is a binary product or not , we explore both possibilities . with monte carlo simulations ,
we investigate the likelihood of finding the observed @xmath182@xmath9 stars @xcite in r136 .
we randomly sample r136-like star clusters for different adopted stellar upper mass limits @xmath183 using the observed imf slope @xcite of @xmath36 , a binary fraction of @xmath0 and that r136 contains about @xmath184 stellar systems more massive than @xmath46 @xcite .
we then compute the average number of stars that are initially more massive than a given mass @xmath185 , @xmath186 , and the probability that at least one star is more massive than @xmath185 , @xmath187 , by repeating each experiment @xmath37 times ( the quoted errors are @xmath151 standard deviations ) .
the average numbers and probabilities for the case that binary interactions did not yet take place are summarised in table [ tab : upper - mass - limit - single - stars ] . for the case that binary interactions already took place
, we assume that all massive binaries with initial periods @xmath188 interact by mass transfer ( which happens within @xmath138@xmath189 ) and that the post - interaction mass is @xmath16 of the total binary mass .
the corresponding average numbers and probabilities for this case can be found in table [ tab : upper - mass - limit - binary - stars ] . in both tables .
[ tab : upper - mass - limit - single - stars ] and [ tab : upper - mass - limit - binary - stars ] , we also give the results for less massive clusters with @xmath190 and @xmath191 stellar systems initially exceeding @xmath46 . through binary mergers , stars of up to @xmath192
can be produced if the star formation process stops at an upper mass of @xmath7 .
however , this scenario requires equal mass o - type binaries which are rare .
we find that with an upper mass of @xmath7 , the probability of forming stars in excess of @xmath193 in r136 is zero of the total system mass , i.e. @xmath194 for @xmath195 . with an upper mass limit of @xmath196 , the probability of forming at least one star of mass @xmath197 increases to @xmath198 , and for an upper mass limit of @xmath199 , the probability of forming at least one star exceeding @xmath193 and @xmath192 is @xmath200 and @xmath201 , respectively .
so , @xmath202 provides a lower limit on the maximum stellar birth mass .
it is also unlikely that the upper mass limit exceeds @xmath203 because then the probability of forming one star above @xmath203 by binary mass transfer increases to @xmath204 but such massive stars are not observed .
we conclude that an upper mass limit in the range of about @xmath11@xmath203 is needed to explain the most massive stars in r136 by binary evolution .
dynamically induced stellar coalescence was proposed as a mechanism to produce the very massive stars in r136 . however , n - body simulations of dynamically induced stellar coalescence typically only produce one to two stars exceeding @xmath202 for r136 adopting an upper mass limit of @xmath7 @xcite , i.e. fewer than observed in r136 . furthermore , the rate of dynamically induced mergers in these simulations should be viewed as an upper limit only .
observational results indeed favour a larger half - mass radius , hence a lower density compared to the simulation assumptions .
similarly , adopting the recent measurements of the orbital distributions of massive binaries further decreases the number of possible dynamical mergers that can overcome the @xmath7 limit by a factor of @xmath205 to @xmath206 .
it appears thus unlikely that dynamically induced stellar coalescence is sufficiently efficient to explain the origin of the very massive stars in r136 if the upper mass limit is @xmath7 . as mentioned above , it is also possible that the four massive stars in r136 were born with their deduced initial masses and did not gain mass by other means .
this provides then an upper limit on the maximum stellar birth mass .
the most massive star found in r136 has an initial mass of @xmath207 @xcite hence , the upper mass limit has to be at least of this order , i.e. @xmath208 .
this initially @xmath9 star allows us to exclude an upper mass limit of @xmath209 with @xmath210 confidence because we expect to find @xmath211 stars that initially exceed @xmath12 in this case while no such star is observed .
however , it becomes more difficult to exclude an upper mass limit of @xmath12 or less because the probability of finding _ no _ star that initially exceeds @xmath203 ( @xmath212 ) is about @xmath213 ; in other words , no star would initially exceed @xmath203 in about every tenth r136-like star cluster for an upper mass limit of @xmath12 .
the probability increases further to @xmath214 for an upper mass limit of @xmath215 .
we conclude that stochastic sampling effects are important even in the richest massive star clusters in the local group .
altogether , we find that current data does not exclude an upper mass limit as high as @xmath216@xmath12 if binary interactions are neglected . however , the most massive star in r136 is a binary product with a probability of @xmath217@xmath50 . including effects of close binary evolution , an initial stellar upper mass limit of at least @xmath202 is required to explain the observed stars with apparent initial masses of about @xmath192 .
the upper mass limit is thus in the range @xmath11@xmath12 , thereby solving the maximum mass problem .
there are several sources of uncertainty that affect theoretical and observed mass functions and hence e.g. our cluster ages derived from them .
it is important to understand the uncertainties to estimate their influence on our conclusions and the derived quantities .
the conclusion that binary effects shape the upper end of the stellar mass function remains unaffected . in sec .
[ sec : uncertainties - models ] , we discuss modelling uncertainties because of the fitting procedure , stellar wind mass loss , binary star evolution and rotation .
observational uncertainties like the influence of different reddening laws on derived stellar masses of stars in the galactic center are discussed in sec .
[ sec : uncertainties - observations ] .
we discuss the influence of dynamical interactions on stellar mass functions in sec .
[ sec : dynamical - effects ] .
star formation histories that are different from single starbursts are considered in sec .
[ sec : sfh ] to investigate whether such scenarios are also consistent with the observed age spread among the most massive stars and the resulting stellar mass functions .
stars in the wind - mass - loss peak of the mass function will very soon leave the main sequence .
the mass of these turn - off stars is a sensitive function of cluster age especially for massive stars , which radiate close to the eddington limit .
massive stars have lifetimes which depend only weakly on mass and hence a small change in age corresponds to a large change in mass .
we can not reproduce the observed mass functions of arches and quintuplet if we change the age of our models in fig .
[ fig : observations ] by more than @xmath218@xmath219 .
we therefore adopt @xmath219 as the age uncertainty associated with our fitting .
the initial binary fraction is best constrained by the number of stars in the mass function tail because it consists only of either post - interaction or pre - interaction , unresolved binaries .
in contrast , the wind - mass - loss peak changes little with the binary fraction .
our observational sample is limited by the exclusion of wnh stars in both arches and quintuplet because no relieable masses of the wnh stars could be determined ( see sec .
[ sec : observations ] .
we thus can not determine the primordial binary fractions accurately , especially because the tail of the quintuplet mass function is not very pronounced .
increasing the age of our quintuplet model by @xmath220 allows for @xmath49 binaries while maintaining a satisfactory , albeit slightly inferior to the best , fit to the mass function .
both clusters are thus consistent with having the same primordial binary fraction .
our wind mass loss prescription slightly underestimates stellar winds compared to the latest predictions .
compared to the most recent stellar evolution models of that use the prescriptions of , we find that our turn - off masses agree to within @xmath221 for initial masses @xmath222 , while in more massive stars our turn - off mass is up to @xmath223 larger , mainly because of the applied wolf - rayet wind mass loss rates in .
the widths of the bins in our model mass functions are @xmath224 , i.e.masses differ by about @xmath15 from bin to bin .
the observed mass functions have bin sizes of @xmath225 , i.e. masses are different by @xmath226 from bin to bin .
wind mass loss prescriptions that lead to stellar masses at the end of the ms that differ by only a few percent result in indistinguishable mass functions our mass functions and conclusions are therefore essentially independent of whether the empirical wind mass loss prescription of or the theoretical prescription of are used .
augmenting our wind loss rate by @xmath0 , we find that an initially @xmath227 star has a turn - off mass of about @xmath228 which matches the recent stellar models by ( compared to @xmath229 in our standard model ) . with the enhanced wind mass loss rate , the arches wind - mass - loss peak corresponds to an initially @xmath230 star with a main - sequence lifetime of @xmath231 , compared to @xmath232 and @xmath114 respectively in our standard model .
the quintuplet wind - mass - loss peak comes from initially @xmath233 stars for which the uncertainty in wind mass loss is @xmath234 .
our quintuplet age estimate is thus robust with respect to the wind mass loss rate uncertainty .
our understanding of binary star evolution in general is subject to uncertainties .
uncertainties that directly influence the shape of the mass function tails are discussed in @xcite .
a further , more quantitative discussion of uncertainties in binary star evolution is found in @xcite and @xcite . here
, we restrict ourselves to ms stars , i.e. to mergers of two ms stars and mass transfer onto ms stars .
mergers that involve a post - ms star form a post - ms object and are thus not considered here .
we assume that two ms stars merge if the mass ratio of the accretor to donor star is less than @xmath235 .
this threshold is calibrated against the detailed binary models of and is of limited relevance to our results : if a binary does not merge but instead transfers mass ( or vice versa ) , the accretor becomes massive because the mass transfer efficiency of ms stars is high . in either case
, the mass gainer will be a massive star @xcite .
the expected binary fraction of stars in the tail of the mass functions however changes : a lower critical mass ratio leads to fewer ms mergers and hence to a higher binary fraction and vice versa . the amount of rejuvenation of ms mergers is determined by the amount of mixing of fresh fuel into the core of the merger product and determines by how much the lifetime of the merger product is prolonged .
the more rejuvenation , the longer the remaining ms lifetime and the more mergers are expected to be found .
we assume that a fraction of @xmath15 of the envelope is mixed into the core , resulting into fairly short remaining ms lifetimes of the merger products compared to the assumption of complete mixing used in the original @xcite code .
recent simulations of massive mergers seem to support the mild mixing as used in our work ( * ? ? ?
* and references therein ) .
the mass transfer efficiency is important for our results .
the more of the transferred mass is accreted during rlof , the larger the final mass of the accreting star .
the maximum reachable mass of any accretor is given by the total mass of the binary ( i.e. at most twice the mass of the donor star ) and the larger the overall mass transfer efficiency , the more binary products exceed the most massive genuine single star . in our models , we limit the mass accretion rate to the thermal timescale of the accretor , which results in higher mass transfer efficiencies the larger the mass ratio and the closer the binary ( see * ? ? ? * ) .
this idea is motivated by detailed binary evolution models .
the initial distributions of primary masses , mass ratios and orbital separations determine the relative fraction of stars that will merge , transfer mass etc .
it turns out that the distribution of orbital separations influences the incidence of binary products most @xcite because initially close binaries transfer mass on average more efficiently than wider binaries .
a distribution of initial orbital separations that favours close binaries therefore leads to on average more massive binary products than distribution functions that favour initially wider binaries .
a more quantitative assessment of the above quoted uncertainties in binary evolution and initial binary distribution functions reveals that a population of ms stars with luminosities @xmath236 ( i.e. o- and b - stars ) contains about @xmath237 binary products if continuous star formation is assumed @xcite .
all in all , depending on the exact assumptions regarding binary evolution , there will be more or fewer stars in the tail of the mass function .
however , the peak - tail structure never disappears unless it is assumed that neither ms mergers nor rlof are able to increase stellar masses which is unphysical . mixing induced by stellar
rotation increases the fuel available to a star and increases its lifetime .
the amount of mixing grows with increasing mass , increasing rotation rate and decreasing metallicity and may contribute to the observed age spreads and mass function tails in arches and quintuplet .
the models of show that the ms lifetime of a @xmath238 star lengthens by @xmath239 and @xmath240 for initial rotational velocities of @xmath241 and @xmath242 , respectively . assuming that the present day distribution of rotational velocities of galactic o- and b - stars approximately represents the initial distribution , no more than @xmath15 and less than @xmath243 of stars would have initial rotation rates exceeding @xmath241 and @xmath244 , respectively , and
are thus expected to be influenced significantly by rotational mixing ( see table 2 in * ? ? ?
* and references therein ) .
this is small compared to the @xmath245 of all o - stars that undergo strong binary interaction during their main - sequence evolution @xcite .
the present day distribution of rotational velocities is probably altered e.g. by binary star evolution such that some of the fast rotators are expected to have gained their fast rotation by binary interactions ( @xcite , but see also @xcite ) . in this respect ,
the expected fraction of genuine galactic single stars that are significantly affected by rotational mixing is even smaller than the above quoted fractions .
the effect of rotation is thus only of limited relevance to our results compared to binary star interactions .
there are two steps involved in determining stellar masses from photometric observations that contribute to the uncertainties of the derived individual stellar masses .
the first step involves the conversion of the observed apparent magnitudes ( fluxes ) to absolute magnitudes and luminosities , respectively , taking amongst others the distance and extinction into account .
the second step involves the conversion of luminosities to stellar masses .
this step relies upon mass - luminosity relations that depend on ( in general a priori unknown ) stellar ages and the applied stellar models . in this section
, we estimate the uncertainty on individual stellar masses introduced by these two steps for stars in arches and quintuplet .
once we know the uncertainties , we can apply them to the turn - off masses derived from the wind mass - loss peak in the mass function to find the corresponding uncertainty in cluster age . in the upper panel of fig .
[ fig : mass - uncertainty ] we show main - sequence mass - luminosity relations of milky way stars of different ages as used in our code ( based on * ? ? ?
not knowing the exact age of a star , but only a probable age range ( here @xmath246 ) , introduces an uncertainty , @xmath247 , on the derived individual stellar masses ( cf . lower panel in fig .
[ fig : mass - uncertainty ] ) .
the more evolved a star and the more massive , the larger is the uncertainty . if we additionally include the uncertainty in the luminosity ( here @xmath248 ; see below ) the uncertainty in the stellar mass grows to @xmath249 .
the uncertainty in luminosity is the dominant contribution here . , and by the combined effect of unknown stellar ages and uncertain luminosities , @xmath250 .
the upper panel illustrates how we estimate these uncertainties for a star with luminosity @xmath251 using the main - sequence mass - luminosity relations of different ages from our code @xcite .
the lower panel shows the uncertainties as a function of mass.,scaledwidth=46.0% ] the uncertainties presented in fig .
[ fig : mass - uncertainty ] are tailored to stars in the arches cluster where stellar masses have been derived by @xcite using a @xmath64 isochrone ( mass - luminosity relation ) whereas our analysis reveals an age of about @xmath114 hence the @xmath246 age range for the mass - luminosity relations .
deriving individual stellar masses using two different extinction laws , find that stellar masses can deviate by up to @xmath128 when either a @xcite extinction law or the traditional @xcite law towards the galactic center are used .
as the appropriate extinction law towards the galactic center line of sight is still a matter of debate , we adopt uncertainties of @xmath248 on luminosities to be in line with the work of .
the situation is similar in quintuplet : we find slightly smaller uncertainties on derived stellar masses for uncertainties of @xmath248 on luminosities and an age range of @xmath252 than given in fig .
[ fig : mass - uncertainty ] . to be conservative , we adopt that individual stellar masses are uncertain by @xmath127 ( i.e. @xmath253 ) in both clusters , which agrees with the diversity of mass estimates in the literature for stars in arches .
find stellar masses that are up to @xmath128 less than those of our analysis for a @xcite extinction law .
an independent study of the arches present - day mass function @xcite shows a similar peak - tail structure at the high mass end using a different extinction law and the derived masses are comparable to those of our analysis @xcite . masses in excess of @xmath7 i.e. larger than our masses have been suggested for the most luminous stars in the arches cluster by @xcite .
hence , an uncertainty of @xmath127 applied to our adopted stellar masses covers the complete range of suggested masses for stars in arches .
the overall structure of the mass functions of arches and quintuplet a stellar - wind peak and binary tail is robust to the mentioned uncertainties .
we also find the peak - tail structure in the mass function of and in mass functions constructed from the photometric data of @xcite and , respectively , using isochrones of different ages .
the whole mass functions shift in mass , but the relative structure stays the same .
the reason is that different extinction laws or isochrones systematically influence all stars in a similar way and do not introduce differential effects . applying @xmath127 uncertainties on stellar masses
leads to turn - off masses of the arches and quintuplet of about @xmath254 and @xmath255 , with associated age uncertainties of @xmath256 and @xmath126 respectively .
the observed present - day mass functions are influenced by dynamical cluster evolution ( * ? ? ?
* and references therein ) .
flat mass function slopes of the order of @xmath257 , compared to a salpeter imf slope of @xmath258 , are likely a consequence of mass segregation in which massive stars sink towards the cluster center for the arches cluster , show that a dynamical model with a standard salpeter imf explains the steepening of the imf slope towards larger distances from the cluster center . while the conclusion is more elusive for the dispersed quintuplet population , the similarity of the mass function slopes in the inner regions of both clusters and the older age of quintuplet suggest that similar processes have shaped the present - day mass function of this cluster as well .
we further investigate whether dynamical interactions in star clusters can also cause the peak in the observed mass functions .
mass segregation flattens the high - mass end of the mass function without producing a peak @xcite while dynamical ejection of stars works on all stars with an ejection efficiency monotonically increasing with mass @xcite as confirmed by the observed galactic fraction of runaway o - stars which is larger than that of runaway b - stars @xcite .
such a smoothly increasing ejection efficiency does not create a peak in the mass function but can give rise to a tail which is however not seen in the mass functions of n - body simulations of the arches cluster @xcite . star formation is not an instantaneous process but lasts a finite amount of time .
observations show that there is an empirical relationship between the duration of star formation and the crossing time of star clusters @xcite .
for example , in westerlund 1 and ngc 3603yc the age difference among stars less massive than about @xmath259 and @xmath260 shows that they were formed within at most @xmath261 and @xmath220 @xcite , respectively , which compare well to the cluster crossing times of @xmath262 and @xmath220 @xcite .
the core of the arches cluster has a radius of about @xmath263 @xcite and a velocity dispersion of @xmath264 @xcite at a distance of @xmath265 corresponding to a crossing time , and hence a duration of star formation , of about @xmath266 .
a similar estimate for the quintuplet cluster is more uncertain because the core radius and especially the velocity dispersion are not well known .
if we assume that the observed central region of quintuplet with a radius of about @xmath267 corresponds to the core radius and that the velocity dispersion is about @xmath268 , the crossing time is @xmath269 .
the mass functions produced by such short periods of star formation are indistinguishable from an instantaneous starburst . the apparent age spread among the most luminous stars in arches and quintuplet is about @xmath20@xmath270 and hence much larger than the estimated star formation periods . in appendix
[ sec : sfh - cont ] we investigate whether a star formation history that is different from a single starburst can also explain the observed peak - tail structure in the mass functions of arches and quintuplet .
we find that this is possible with e.g. a two - stage starburst but the age spread among the most massive stars would then be inconsistent with observations .
massive stars rapidly change their mass , thereby altering the stellar mass function .
stellar wind mass loss reduces stellar masses such that stars accumulate near the high mass end of present day mass functions , creating a bump whose position reveals the mass of the turn - off stars and hence the age of young star clusters @xcite .
binary stars are frequent and important for massive star evolution because of binary mass exchange @xcite : mass transfer and stellar mergers increase stellar masses and create a tail of rejuvenated binary products at the high mass end of mass functions @xcite .
we model the observed mass functions of the young arches and quintuplet star clusters using a rapid binary evolution code to identify these two features and to address two pressing controvercies : 1 . _
the cluster age problem : _ the most massive stars in arches and quintuplet , the wnh stars , appear to be younger than the less massive o - stars .
this is not expected from star cluster formation @xcite but well known in older clusters under the blue straggler phenomenon .
the maximum mass problem : _ a stellar upper mass limit of @xmath7 is observationally determined @xcite . in contrast , the supernova sn 2007bi is thought to be a pair - instability supernova from an initially @xmath10 star @xcite and four stars greatly exceeding this limit are found in the r136 cluster in the large magellanic cloud @xcite .
we identify the peak and tail in the observed mass functions of the arches and quintuplet clusters . by fitting our models to the observations , we determine the ages of arches and quintuplet to be @xmath1 and @xmath2 , respectively .
this solves the age problem because the most massive stars in arches and quintuplet are rejuvenated products of binary mass transfer and the ages derived from these stars are therefore significantly underestimated . while our age error bars are still large mostly because of uncertain absolute magnitudes our method removes the ambiguity in the age determination . for the age determination of older star clusters , blue stragglers are eminently disregarded when isochrones are fitted to the turn - off in hertzsprung - russell diagrams .
our analysis shows that for young star clusters , where the higher fraction of interacting binaries produces even more blue stragglers , they obviously need to be disregarded as well in order to derive the correct cluster age . even without modelling the observed mass functions , the ages and also the mass lost by the turn - off stars during their main - sequence evolution can be determined from the mass function alone .
refilling the mass function above the present day mass of the turn - off stars with the number of excess stars in the wind - mass - loss peak gives the initial mass of the turn - off stars and hence the cluster age .
the derived ages agree with the more accurate ages from detailed modelling of the observed mass functions . according to this new method , the turn - off stars in arches lost @xmath106@xmath117 of their initial @xmath109@xmath110 and @xmath118@xmath119 of their initial @xmath111@xmath271 in quintuplet .
this is the first direct measurement of stellar wind mass loss which does not rely on derivations of stellar wind mass loss rates .
monte carlo experiments to investigate the effects of stochastic sampling show that the most massive star in the arches and quintuplet clusters is expected to be a rejuvenated product of binary mass transfer after @xmath3 and @xmath4 , respectively . at their present age ,
the probability that the most massive star in arches and quintuplet is a product of binary mass exchange is @xmath167 and the most massive @xmath168 and @xmath169 stars in arches and quintuplet , respectively , are expected to be such rejuvenated binary products .
our findings have implications for the maximum mass problem . the arches cluster is older than previously thought and its most massive stars are most likely binary products . the mass function is thus truncated by finite stellar lifetimes and not by an upper mass limit . to constrain a potential stellar upper mass limit
, we consider the massive cluster r136 in the large magellanic cloud which is thought to be so young that its initially most massive stars are still alive today .
we find that the most massive star is a binary product with a probability of @xmath272 , depending on the exact , albeit yet uncertain cluster age ( sec .
[ sec : upper - mass - limit ] ) .
assuming binaries already interacted , a stellar upper mass limit of at least @xmath202 is needed to form the observed @xmath8@xmath9 stars in r136 .
it can also not exceed @xmath203 because then the probability of forming stars above e.g. @xmath203 becomes larger than about @xmath273 but such stars are not observed . assuming that no binary interactions changed the masses of the very massive stars in r136 , a stellar upper mass limit of up to @xmath216@xmath12 can not be fully excluded because of stochastic sampling even in this rich star cluster .
the upper mass limit is thus likely in the range @xmath11@xmath12 , thereby solving the maximum mass problem .
we conclude that the most massive stars in the universe may be the rejuvenated products of binary mass transfer . because of their extreme mass and luminosity , radiation feedback from these stars is crucial to observable properties of young stellar populations , to the state of the interstellar medium around young stellar clusters and even to the reionization of the universe after the big bang .
our results have strong implications for understanding star - forming regions nearby and at high redshift as observationally derived fundamental properties like initial mass functions are based on the assumption that the brightest stars are single and less massive than @xmath7 .
these very massive stars are thought to die as pair - instability supernovae and produce huge , so far unaccounted contributions to the chemical enrichment of nearby and distant galaxies and their final explosions may be observable throughout the universe .
understanding the most massive stars in young nearby star clusters is an essential step towards investigating these exciting phenomena which shape our cosmos .
we thank the referee , dany vanbeveren , for carefully reading our manuscript and constructive suggestions .
f.r.n.s . acknowledges the fellowships awarded by the german national academic foundation ( studienstiftung ) and the bonn - cologne graduate school of physics and astronomy .
r.g.i . would like to thank the alexander von humboldt foundation . s.d.m
acknowledges support by nasa through hubble fellowship grant hst - hf-51270.01-a 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 and the einstein fellowship program through grant pf3 - 140105 awarded by the chandra x - ray center , which is operated by the smithsonian astrophysical observatory for nasa under the contract nas8 - 03060 . b.h . and a.s .
acknowledge funding from the german science foundation ( dfg ) emmy noether program under grant sto 496 - 3/1 .
cccccccc & & & & & & & & @xmath274 & @xmath275 & @xmath276 & @xmath187 & @xmath276 & @xmath187 & @xmath276 & @xmath187@xmath277 & @xmath278 & @xmath279 & @xmath280 & @xmath281 & @xmath282 & @xmath283 & @xmath167 & @xmath11 & @xmath284 & @xmath285 & @xmath286 & @xmath287 & @xmath288 & @xmath167 & @xmath289 & @xmath290 & @xmath291 & @xmath292 & @xmath293 & @xmath294 & @xmath167 & @xmath241 & @xmath295 & @xmath296 & @xmath297 & @xmath298 & @xmath299 & @xmath282 & @xmath191 & @xmath300 & @xmath301 & @xmath302 & @xmath303 & @xmath304 & @xmath305 & @xmath216 & @xmath306 & @xmath307 & @xmath308 & @xmath309 & @xmath310 & @xmath311 & @xmath312 & @xmath313 & @xmath314 & @xmath315 & @xmath316 & @xmath317 & @xmath318 & @xmath319 & @xmath320 & @xmath321 & @xmath284 & @xmath322 & @xmath211 & @xmath323@xmath37 & @xmath278 & @xmath324 & @xmath325 & @xmath326 & @xmath167 & @xmath327 & @xmath167 & @xmath11 & @xmath328 & @xmath329 & @xmath330 & @xmath331 & @xmath332 & @xmath167 & @xmath289 & @xmath333 & @xmath334 & @xmath335 & @xmath336 & @xmath337 & @xmath282 & @xmath241 & @xmath300 & @xmath338 & @xmath339 & @xmath340 & @xmath304 & @xmath341 & @xmath191 & @xmath313 & @xmath342 & @xmath166 & @xmath343 & @xmath344 & @xmath345 & @xmath216 & @xmath320 & @xmath346 & @xmath328 & @xmath347 & @xmath348 & @xmath349 & @xmath312 & @xmath350 & @xmath351 & @xmath290 & @xmath352 & @xmath353 & @xmath354 & @xmath319 & @xmath355 & @xmath356 & @xmath295 & @xmath357 & @xmath315 & @xmath358@xmath319 & @xmath278 & @xmath359 & @xmath360 & @xmath361 & @xmath167 & @xmath362 & @xmath167 & @xmath11 & @xmath363 & @xmath364 & @xmath310 & @xmath365 & @xmath366 & @xmath167 & @xmath289 & @xmath367 & @xmath368 & @xmath339 & @xmath369 & @xmath370 & @xmath331 & @xmath241 & @xmath320 & @xmath371 & @xmath372 & @xmath373 & @xmath374 & @xmath323 & @xmath191 & @xmath375 & @xmath376 & @xmath333 & @xmath377 & @xmath315 & @xmath378 & @xmath216 & @xmath379 & @xmath380 & @xmath306 & @xmath381 & @xmath333 & @xmath382 & @xmath312 & @xmath383 & @xmath384 & @xmath355 & @xmath385 & @xmath320 & @xmath386 & @xmath319 & @xmath387 & @xmath388 & @xmath387 & @xmath388 & @xmath387 & @xmath388@xmath216 & @xmath278 & @xmath389 & @xmath390 & @xmath391 & @xmath282 & @xmath392 & @xmath167 & @xmath11 & @xmath295 & @xmath393 & @xmath317 & @xmath394 & @xmath395 & @xmath167 & @xmath289 & @xmath396 & @xmath397 & @xmath398 & @xmath399 & @xmath400 & @xmath336 & @xmath241 & @xmath375 & @xmath401 & @xmath333 & @xmath402 & @xmath359 & @xmath403 & @xmath191 & @xmath383 & @xmath404 & @xmath320 & @xmath405 & @xmath406 & @xmath407 & @xmath216 & @xmath387 & @xmath388 & @xmath387 & @xmath388 & @xmath387 & @xmath388@xmath241 & @xmath278 & @xmath284 & @xmath408 & @xmath409 & @xmath410 & @xmath411 & @xmath167 & @xmath11 & @xmath300 & @xmath412 & @xmath279 & @xmath413 & @xmath414 & @xmath415 & @xmath289 & @xmath375 & @xmath416 & @xmath295 & @xmath417 & @xmath418 & @xmath419 & @xmath241 & @xmath387 & @xmath388 & @xmath387 & @xmath388 & @xmath387 & @xmath388@xmath11 & @xmath278 & @xmath367 & @xmath420 & @xmath421 & @xmath422 & @xmath423 & @xmath287 & @xmath11 & @xmath387 & @xmath388 & @xmath387 & @xmath388 & @xmath387 & @xmath388 cccccccc & & & & & & & & @xmath274 & @xmath275 & @xmath276 & @xmath187 & @xmath276 & @xmath187 & @xmath276 & @xmath187@xmath216 & @xmath278 & @xmath359 & @xmath424 & @xmath425 & @xmath282 & @xmath426 & @xmath167 & @xmath11 & @xmath290 & @xmath427 & @xmath428 & @xmath429 & @xmath430 & @xmath167 & @xmath289 & @xmath300 & @xmath431 & @xmath279 & @xmath432 & @xmath433 & @xmath341 & @xmath241 & @xmath320 & @xmath434 & @xmath435 & @xmath436 & @xmath339 & @xmath437 & @xmath191 & @xmath438 & @xmath439 & @xmath300 & @xmath440 & @xmath435 & @xmath441 & @xmath216 & @xmath383 & @xmath442 & @xmath355 & @xmath385 & @xmath320 & @xmath443 & @xmath312 & @xmath444 & @xmath445 & @xmath438 & @xmath446 & @xmath375 & @xmath416 & @xmath319 & @xmath444 & @xmath447 & @xmath383 & @xmath448 & @xmath379 & @xmath449 & @xmath450 & @xmath451 & @xmath452 & @xmath383 & @xmath453 & @xmath383 & @xmath454 & @xmath455 & @xmath451 & @xmath456 & @xmath444 & @xmath457 & @xmath444 & @xmath458 & @xmath184 & @xmath451 & @xmath459 & @xmath451 & @xmath460 & @xmath451 & @xmath461 & @xmath462 & @xmath387 & @xmath463 & @xmath387 & @xmath464 & @xmath387 & @xmath463 & @xmath465 & @xmath387 & @xmath388 & @xmath387 & @xmath388 & @xmath387 & @xmath388 & @xmath466 & @xmath387 & @xmath388 & @xmath387 & @xmath388 & @xmath387 & @xmath388@xmath191 & @xmath278 & @xmath166 & @xmath467 & @xmath468 & @xmath167 & @xmath469 & @xmath167 & @xmath11 & @xmath470 & @xmath471 & @xmath472 & @xmath293 & @xmath473 & @xmath282 & @xmath289 & @xmath396 & @xmath474 & @xmath166 & @xmath475 & @xmath476 & @xmath477 & @xmath241 & @xmath355 & @xmath478 & @xmath333 & @xmath479 & @xmath480 & @xmath481 & @xmath191 & @xmath383 & @xmath482 & @xmath375 & @xmath483 & @xmath306 & @xmath204 & @xmath216 & @xmath444 & @xmath484 & @xmath379 & @xmath485 & @xmath350 & @xmath486 & @xmath312 & @xmath451 & @xmath487 & @xmath383 & @xmath482 & @xmath438 & @xmath488 & @xmath319 & @xmath451 & @xmath489 & @xmath444 & @xmath490 & @xmath383 & @xmath491 & @xmath450 & @xmath451 & @xmath459 & @xmath451 & @xmath452 & @xmath444 & @xmath457 & @xmath455 & @xmath387 & @xmath388 & @xmath387 & @xmath388 & @xmath387 & @xmath388 & @xmath184 & @xmath387 & @xmath388 & @xmath387 & @xmath388 & @xmath387 & @xmath388 & @xmath462 & @xmath387 & @xmath388 & @xmath387 & @xmath388 & @xmath387 & @xmath388@xmath241 & @xmath278 & @xmath389 & @xmath492 & @xmath493 & @xmath167 & @xmath494 & @xmath167 & @xmath11 & @xmath495 & @xmath496 & @xmath497 & @xmath336 & @xmath286 & @xmath287 & @xmath289 & @xmath320 & @xmath498 & @xmath435 & @xmath436 & @xmath339 & @xmath499 & @xmath241 & @xmath500 & @xmath501 & @xmath350 & @xmath502 & @xmath306 & @xmath503 & @xmath191 & @xmath383 & @xmath504 & @xmath438 & @xmath505 & @xmath350 & @xmath506 & @xmath216 & @xmath444 & @xmath507 & @xmath383 & @xmath508 & @xmath379 & @xmath509 & @xmath312 & @xmath451 & @xmath461 & @xmath444 & @xmath510 & @xmath511 & @xmath512 & @xmath319 & @xmath451 & @xmath513 & @xmath451 & @xmath456 & @xmath451 & @xmath460 & @xmath450 & @xmath387 & @xmath388 & @xmath387 & @xmath388 & @xmath387 & @xmath388 & @xmath455 & @xmath387 & @xmath388 & @xmath387 & @xmath388 & @xmath387 & @xmath388@xmath289 & @xmath278 & @xmath514 & @xmath515 & @xmath516 & @xmath287 & @xmath517 & @xmath167 & @xmath11 & @xmath306 & @xmath518 & @xmath519 & @xmath520 & @xmath310 & @xmath365 & @xmath289 & @xmath500 & @xmath501 & @xmath320 & @xmath521 & @xmath295 & @xmath522 & @xmath241 & @xmath511 & @xmath523 & @xmath438 & @xmath524 & @xmath525 & @xmath434 & @xmath191 & @xmath451 & @xmath452 & @xmath383 & @xmath384 & @xmath379 & @xmath509 & @xmath216 & @xmath451 & @xmath459 & @xmath451 & @xmath526 & @xmath444 & @xmath527 & @xmath312 & @xmath387 & @xmath388 & @xmath387 & @xmath388 & @xmath387 & @xmath388 & @xmath319 & @xmath387 & @xmath388 & @xmath387 & @xmath388 & @xmath387 & @xmath388@xmath11 & @xmath278 & @xmath528 & @xmath402 & @xmath529 & @xmath429 & @xmath169 & @xmath167 & @xmath11 & @xmath379 & @xmath530 & @xmath300 & @xmath531 & @xmath532 & @xmath533 & @xmath289 & @xmath383 & @xmath534 & @xmath355 & @xmath535 & @xmath320 & @xmath321 & @xmath241 & @xmath451 & @xmath536 & @xmath383 & @xmath537 & @xmath500 & @xmath201 & @xmath191 & @xmath387 & @xmath463 & @xmath387 & @xmath464 & @xmath387 & @xmath464 & @xmath216 & @xmath387 & @xmath388 & @xmath387 & @xmath388 & @xmath387 & @xmath388@xmath278 & @xmath278 & @xmath355 & @xmath538 & @xmath333 & @xmath539 & @xmath315 & @xmath343 & @xmath11 & @xmath383 & @xmath540 & @xmath355 & @xmath541 & @xmath320 & @xmath521 & @xmath289 & @xmath451 & @xmath542 & @xmath451 & @xmath526 & @xmath444 & @xmath543 & @xmath241 & @xmath387 & @xmath388 & @xmath387 & @xmath388 & @xmath387 & @xmath388
here , we investigate changes in the mass function due to a star formation scenario which deviates from a true starburst in order to understand whether the observed mass functions of arches and quintuplet can be reproduced without binaries .
we analyse two scenarios : ( a ) a period of prolonged but constant star formation rate and ( b ) two instantaneous starbursts separated in time .
the latter scenario ( b ) not only represents a two stage starburst within one cluster but also two merged star clusters where stars in each cluster are coeval .
this situation most probably applies to the massive star cluster r136 in the large magellanic cloud which is thought to be a double cluster in the process of merging @xcite .
we compute mass functions for the star formation scenarios ( a ) and ( b ) ( which include single , true starbursts ) and search for parameter values that minimise the least - square deviation , of the modelled ( @xmath544 ) from the observed ( @xmath545 ) mass functions of the arches and quintuplet clusters assuming poisson uncertainties , i.e. @xmath546 , @xmath547 where @xmath94 is the number of mass bins .
exemplary star formation scenarios are described in table [ tab : sf - models ] and the resulting mass functions are compared to observations in fig .
[ fig : sfh ] . among these examples
are those star formation scenarios that lead to the best fits ( models a2 , a4 , q2 and q4 ) .
ccccccp9 cm sf model & @xmath548 & @xmath549 & @xmath150 & @xmath550 & @xmath551 & description a1 & & & & @xmath552 & @xmath553 & power law mass function truncated at the most massive observed star ; power law index @xmath42a2 & @xmath205 & & @xmath49 & @xmath554 & @xmath555 & single starburst at @xmath556a3 & @xmath557 & @xmath558 & @xmath559 & @xmath560 & @xmath561 & constant sf between @xmath556 and @xmath562a4 & @xmath563 & @xmath558 & @xmath559 & @xmath564 & @xmath565 & two starbursts at @xmath556 and @xmath562q1 & & & & @xmath566 & @xmath567 & power law mass function truncated at the most massive observed star ; power law index @xmath42q2 & @xmath568 & & @xmath50 & @xmath235 & @xmath555 & single starburst at @xmath556q3 & @xmath568 & & @xmath559 & @xmath569 & @xmath570 & single starburst at @xmath556q4 & @xmath571 & @xmath572 & @xmath559 & @xmath573 & @xmath574 & constant sf between @xmath556 and @xmath562 from table [ tab : sf - models ] and the top panels of fig .
[ fig : sfh ] it is evident that simple power - law mass functions ( models a1 and q1 ) fit the observed mass functions of arches and quintuplet much worse than the best single starburst models including binary stars ( a2 and q2 ) .
especially the mass region around the wind - mass - loss peak is not fitted well by models a1 and q1 ( see @xmath575 in table [ tab : sf - models ] ) .
the mass functions of the arches and quintuplet clusters do not follow simple power laws . in the bottom panels of fig .
[ fig : sfh ] , we also present models of the observed mass functions of the arches and quintuplet clusters _ without _ binary stars
. we do not find satisfactory models which fit peak and tail simultaneously with a single starburst without binaries .
model a3 for example fits the peak due to mass - loss well ( @xmath576 ) but fails to explain the high mass end of the observed mass function ( @xmath577 ) .
we can improve this situation by adding a younger stellar population that fits the tail .
such a scenario is given by model a4 which fits the peak and the total mass function .
however , this two - component model has an age spread of @xmath578 which is more than twice as large as the observed age discrepancy of about @xmath579 in the arches cluster and much larger than the expected star formation duration .
a two component solution is not needed to model the observed mass function of the quintuplet cluster because the tail of the mass function is not very pronounced .
consequently our models q3 and q4 predict no or a too small age spread contrary to the observations .
the age spread of @xmath239 of model q4 might be compatible with the above estimated star formation duration given the quite uncertain core radius and velocity dispersion of quintuplet .
the single starburst model q3 is shown to illustrate the difference between the mass functions with ( q2 ) and without ( q3 ) binaries .
the tail of the mass function is however underestimated in the observed mass function in fig .
[ fig : sfh ] because no self - consistent mass determination for the three wnh stars in the core of quintuplet is available .
if the tail were visible , we could of course model it by an additional younger population as is in arches .
in summary , we conclude that we can reproduce the mass functions of arches and quintuplet without binaries but with freedom in the star formation history
. however , the best fit star formation parameters ( e.g. the age spread of @xmath578 ) are inconsistent with other observables .
our single starburst models which include binaries are thus the only models which fulfil all observational constraints .
these models are also consistent with a star formation duration of the order of the crossing time of the cluster . | massive stars rapidly change their masses through strong stellar winds and mass transfer in binary systems .
the latter aspect is important for populations of massive stars as more than @xmath0 of all o - stars are expected to interact with a binary companion during their lifetime .
we show that such mass changes leave characteristic signatures in stellar mass functions of young star clusters which can be used to infer their ages and to identify products of binary evolution .
we model the observed present day mass functions of the young galactic arches and quintuplet star clusters using our rapid binary evolution code .
we find that shaping of the mass function by stellar wind mass loss allows us to determine the cluster ages to @xmath1 and @xmath2 , respectively .
exploiting the effects of binary mass exchange on the cluster mass function , we find that the most massive stars in both clusters are rejuvenated products of binary mass transfer , i.e. the massive counterpart of classical blue straggler stars .
this resolves the problem of an apparent age spread among the most luminous stars exceeding the expected duration of star formation in these clusters .
we perform monte carlo simulations to probe stochastic sampling , which support the idea of the most massive stars being rejuvenated binary products .
we find that the most massive star is expected to be a binary product after @xmath3 in arches and after @xmath4 in quintuplet .
today , the most massive @xmath5 stars in arches and @xmath6 in quintuplet are expected to be such objects .
our findings have strong implications for the stellar upper mass limit and solve the discrepancy between the claimed @xmath7 limit and observations of fours stars with initial masses of @xmath8@xmath9 in r136 and of sn 2007bi , which is thought to be a pair - instability supernova from an initial @xmath10 star . using the stellar population of r136 ,
we revise the upper mass limit to values in the range @xmath11@xmath12 . |
dilute bose - einstein condensates ( becs ) are produced and probed using atomic physics tools @xcite , and their connection to condensed - matter physics is most clear when studying collective excitations @xcite , as well as the sound propagation @xcite .
much of our understanding of the nature of sound in quantum fluids comes from studies carried out in liquid helium @xcite .
recently , with gaseous condensates , the researchers have turned to these decades - old theories from condensed matter physics to test their validity thoroughly and very successfully @xcite .
nowadays , these versatile experimental system are even being used as fast complex hamiltonian solvers , which may provide `` on demand '' solutions to particular interesting problems in condensed matter physics .
the nature of collective excitations in a homogeneous bose gas depends on the hierarchy of three length scales : the excitation reduced wavelength , @xmath1 ; the healing length , @xmath2 ; and the mean free path for collisions , @xmath3 . in a inhomogeneous bose trapped gas ,
the excitations are influenced by an additional length : the typical condensate size , the thomas - fermi radii .
this gives rise to three different excitation regimes : hydrodynamic , collision and collisionless .
the first observations of the low - lying excitations in gaseous becs were made at jila @xcite and at the mit @xcite in the zero temperature limit .
soon after , stringari carried on in depth theoretical studies of the collective modes @xcite creating the base framework to understand the observations . in this work ,
we investigate the collective excitations of @xmath0rb condensates held in a cigar - shaped quadrupole - ioffe configuration ( quic ) trap @xcite .
the collective excitations of the condensate are produced by an additional , time varying , magnetic field gradient superimposed to a quic potential .
besides the expected oscillation of the center of mass , we observed the bec undergoing shape oscillations .
the frequency of the observed collective excitations were measured and compared to the theoretical predictions .
we see evidence of a non - linear dependence on the amplitude of the perturbing field gradient , not yet predicted by the theory .
we believe that the study of collective modes is relevant because they present the fundamental features of quantum degenerate gases .
furthermore , we would like to better understand the vortex nucleation and the onset of turbulence in perturbed condensates . in the next sections , we present a brief discussion on the collective modes observed in becs , followed by a description of the experimental setup used to acquire the data .
the results are then presented and discussed based on the current theory for the dispersion relations , where good agreement is found for the measured mode frequencies .
finally , the conclusions are presented .
bose - einstein condensates were object of intensive studies early since they were first produced in gaseous alkali metal samples .
the low - lying collective excitations were theoretically investigated and predicted just one year after the initial experimental bec observations @xcite . back then , stringari @xcite was able to derive analytic solutions for both isotropic and anisotropic harmonic potentials in the thomas - fermi regime .
the normal modes of a bec are classified by quantum numbers @xmath4 , where @xmath5 is the radial quantum number and @xmath6 stand for the total angular momentum and its axial projection quantum numbers , respectively . for cylindrical symmetry ( as in our trap ) , @xmath7 is still a good quantum number , but @xmath8 is not .
thus , the normal modes are a superposition of wave functions with the same @xmath7 .
stringari discussed an anisotropic harmonic potential with axial symmetry and showed that the lowest @xmath9 modes are coupled excitations of the @xmath10 modes , quadrupolar surface oscillation , and also the @xmath11 symmetries . for a cigar - shaped condensates
the lowest modes were predicted at frequencies @xcite : @xmath12 @xmath13 and @xmath14 in the above equations @xmath15 , @xmath16 are the axial , radial trap frequencies which will be properly defined in the next section , and @xmath17 is the ratio of those two frequencies known as the asymmetry parameter . in the large anisotropy limit @xmath18 , the solutions of eq .
[ eq3 ] above are : @xmath19 and @xmath20 . the first , low - lying , shape oscillations associated with the above dispersion relations are : 1 ) the cylindrically symmetric ( slow ) @xmath9 quadrupole mode where the axial ( @xmath21axis ) length oscillates out of phase with the radial lengths ; 2 ) the in - phase ( fast ) , mostly radial , @xmath22 breathing mode ; and 3 ) the non - symmetric radial oscillation @xmath23 quadrupole mode where the two radial lengths oscillate out of phase .
the experimental sequence to produce the bec runs as follows .
first , @xmath24 becs containing about @xmath25 atoms are formed in the @xmath26 hyperfine state , with a small thermal fraction in a cigar - shaped quic magnetic trap . in order to achieve the degeneracy ,
the atoms undergo forced evaporation for @xmath27 , going from @xmath28 down to about @xmath29 , following a non analytic curve experimentally determined by optimizing the phase space density and the runaway condition in each step .
we typically produce samples with no more than 30 or 35% of thermal fraction , well into the thomas - fermi regime .
the measured trapping frequencies are : @xmath30 in the symmetry axis , and @xmath31 in the radial direction .
the radial trapping frequency depends on the offset field , @xmath32 , at the trap bottom and the magnetic field gradient in the radial direction , which can be written as : @xmath33 where @xmath34 , @xmath35 and @xmath36 are the magnetic quantum number , the bohr magneton , and land g - factor for the hyperfine state @xmath37 , and @xmath7 is the atomic mass .
the axial trapping frequency is determined by the axial curvature @xmath38 through @xmath39 .
the values of @xmath32 , @xmath40 , and @xmath38 are determined based on the geometric parameters of the quic coils and the electric current running through them . for our system ,
the calculated values are @xmath41 , @xmath42 , @xmath43 , and @xmath44 .
correspondingly , the theoretical results of the radial and axial trapping frequencies are @xmath45 and @xmath46 , respectively , for a current of @xmath47 through the coils .
finally , we have @xmath48 , @xmath49 , and the anisotropy parameter is @xmath50 .
to induce the collective excitations , an additional sinusoidal magnetic field gradient is turned on and superimposed on the atoms just after the bec is produced .
the external gradient is a quadrupolar field generated by an anti - helmholtz coil pair set with symmetry axis closely parallel to the weak trap axis .
the coil pair is not perfectly aligned to the condensate axis ( @xmath51 ) , so its field gradient , near the trapping region , has components parallel to each trap eigen - axis .
a superposition of shape morphing , trap bottom displacement and axial rotation is coupled to the bec .
additional , sinusoidal field gradients going from zero up to @xmath52 , along the vertical axis , were applied to the trapped atoms inducing the onset of the collective modes .
the extra gradient oscillating at @xmath53 was kept on for six extra periods of trap oscillation and then switched - off . in fig.[fig1 ] , we present a typical set of the perturbed bec samples imaged for different hold times , nearly completing a full oscillation cycle .
[ ht ! ]
we then kept the bec held in the trap for different time intervals , @xmath54 , ranging from @xmath55 to @xmath56 after which we observe it by performing a resonant time - of - flight ( tof ) absorption imaging , after @xmath57 free fall to avoid pixel saturation and increase the accuracy .
center - of - mass position of the cloud as well as its axes sizes and orientations are recorded and plotted as a function of the @xmath54 , from where we extract both frequency and amplitude of the collective modes .
using the technique explained in the previous section , we are able to observe dipolar , quadrupolar , breathing and also the scissors modes in our becs . for small excitation amplitudes , we observed the tilting about the gravity direction ( @xmath58 ) .
such periodic axial tilting is generated by a sudden rotation of the confining trap , and it is known in the literature as scissors mode .
it was theoretically studied by gury - odelin @xcite and others and will not be discussed here . here , we will focus on the dipolar and the quadrupolar modes only . after being perturbed , the bec center - of - mass always oscillate in the dipolar mode @xcite , into the trap .
this mode corresponds to the full oscillation at the trap frequency in each direction .
we have also observed the onset of the quadrupolar mode @xcite .
since the longitudinal and transverse directions do not oscillate exactly at @xmath59 out - of - phase , we believe that the higher order modes may also be onset .
moreover , the condensate s long axis cyclic tilting was observed and priviously reported @xcite , see fig.[fig1 ] .
the observation of the scissors mode @xcite shows that the perturbing field gradient is able to cause the bec to rotate .
[ ht ! ]
the study of the low - lying collective excitations was carried on by following the center - of - mass ( cm ) motion and the aspect ratio ( ar ) of the becs as a function of the hold time . for small amplitudes ,
the cm is usually located near the center of the atomic density distribution , which is also the geometric center of the bec .
but , in some situations , at large perturbing amplitudes ( @xmath60 ) , we observed slosh dynamics taking place , clearly displacing the thermal and condensed fractions in opposite , out - of - phase , cm motion @xcite .
it is important to mention that the dipolar mode preserves the ar when compared to an unperturbed bec .
therefore , the ar analysis are used only to study the higher order modes ( quadrupolar and breathing ) . in fig.[fig2](a )
, we present the results acquired by directly tracking the center - of - mass ( cm ) motion as a function of the hold time for four different perturbing amplitudes .
one may note that , the perturbing energy not only increases the cm displacement from the rest position , but also shifts the phase of the sinusoidal motion .
fig.[fig2](b ) presents the ar evolution with the hold time , measured for the quadrupolar mode .
again , note the phase shift introduced in the ar evolution as the perturbing energy amplitude raises from @xmath61 ( 10% ) to @xmath62 ( 70% ) .
the data shown in figs.[fig2](a ) and ( b ) are fitted by simple sinusoidal functions from where we determine the oscillation frequency of the collective modes and their amplitudes .
the evolution of the cm and the ar as a function of the perturbing amplitude is shown in fig.[fig3](a ) .
on the one hand , the cm oscillation amplitude proportionally increases with the excitation amplitude starting from zero , as expected . on the other hand ,
the evolution of the ar clearly show different evolution slopes , depending on perturbing amplitude range . and
it displays an interesting non - linear dependence near @xmath63 .
a small region around @xmath64 presents a interesting non - linear behavior , rapidly evolving from the initial to the final ( steeper ) slope , strongly suggesting the emergence of a different regime for perturbing amplitudes lager than @xmath63 .
we do not have yet a theoretical explanation for this observation .
fig.[fig3](b ) presents the changes on the dipolar and quadrupolar oscillation frequencies as the perturbing amplitude increases .
we measured @xmath65 for the dipolar oscillations , compared to @xmath16 predicted . and , @xmath66 for the fast oscillation , quadrupolar mode , which is to be compared to the theoretical prediction : @xmath20 .
the observed modes were previously studied and analytic expressions for the dispersion relations were theoretically derived @xcite .
the agreement found between our results and the predicted frequencies , eqs.[eq1 ] to [ eq3 ] , presented in section 2 is very good .
the results show almost no change with the frequency for amplitudes up to about @xmath63 and a slight decrease ( @xmath67 ) for larger amplitudes .
similar observations were seen before @xcite .
we expected to see a constant evolution along the full range and believe the slight change may attributed to nonlinear interactions coupling the condensate normal modes @xcite . for nearly pure condensates , the thermal excitations should have no effect at all and , to the best of our knowledge , there is still no theoretical explanation for a possible slow down of the collective modes frequency of a trapped bec .
[ ht ! ]
we believe though , that the different behavior observed in between the cm oscillation and the shape oscillation ( quadrupolar ) might be due to way we perturb the bec .
the addition of a small amplitude , time - varying , tilted magnetic field gradient will mostly displace the cm of the bec .
but , once the axial and the radial trapping gradients are a factor of two different , it is enough to induce rotation and shape oscillations .
this observation is in agreement with results presented in figs.[fig1 ] - [ fig3 ] .
therefore , small perturbation results in cm displacement with little effect on the bec shape .
as the perturbation amplitude increases , the condensate is pushed farther away from the trap bottom . in a quic trap
, the bec needs to be displaced away from the trap center , up to regions with steeper field curvature ( see fig.2(d ) in the original paper @xcite ) , where the unbalanced compression is likely to be effective inducing shape oscillations .
that is consistent with the near resonant perturbation used to drive our bec .
moreover , the quic traps are known to be asymmetric along the weak axis direction due to the presence of a single ioffe coil , making it easier for the condensate to reach the non symmetric boundaries of the trapping potential .
this also seems consistent with the observation of a change in the curvature shown in fig.[fig3 ] , where the bec breaks free from the inner harmonic ( symmetric ) trapping region and travels through the outermost , anharmonic , trap region .
these outer trap regions present much stepper field gradient and curvature making them suitable to produce shape oscillations in a bec by unbalanced compression . beyond that
, we believe that the steep gradient regions might be the inducing mode coupling and and even the dephasing observed in fig.[fig2 ] .
this complex coupling dynamics seems particular to our experimental system .
it may also be the key to the better understand the vortex nucleation and the turbulence emergence observed in previous experiments @xcite . in contrast , the presented results are in good agreement with the theoretical predictions except for the observed threshold for the quadrupolar shape oscillations .
we have studied the excitations induced in cigar - shaped condensates produced in a quic trap , near zero temperature .
collective modes were excited by sinusoidally varying currents running through additional coils , externally set for this purpose .
the excitation was not axially symmetric , and so the @xmath68 modes were excited too .
dipolar oscillations were readily induced and used to accurately measure the trap frequencies .
at least , two additional shape oscillations were set on and studied . the higher frequency ,
primarily , radial breathing mode was excited and its frequency was measured , in very good agreement with the predicted theoretical value .
finally , we have observed an interesting , nonlinear , onset dependence on the minimum excitation amplitude needed to start changing aspect ratio . | we report the experimental observation of the collective excitations induced in a magnetically trapped @xmath0rb bose - einstein condensate .
low - lying mode excitations were studied by tracking the condensate s center - of - mass displacement , and its aspect ratio as a function of the hold time in the trap .
we were able to partially control the modes onset by modulating the amplitude of the additional field gradient used to excite the bec .
the measured excitation frequencies were found to be in good agreement with the literature .
we have also found that the modulation amplitude was able to change the phase of the center - of - mass oscillation . finally , an interesting , non - linear dependence was observed on the condensate aspect ratio as a function of the perturbing amplitude which induces the quadrupolar mode . |
the presence of a diffuse extended low density warm ionized medium(eldwim ) within the inner galaxy has been established by low frequency(@xmath32ghz ) radio recombination line(rrl or rl for short ) observations(lockman 1976 ; mezger 1978 ; ananatharamaiah 1985 ) .
this component is understood to have a typical electron density @xmath4 of @xmath5 at temperatures of @xmath6 k. the eldwim was first considered by mezger(1978 ) who called it as the `` extended low density fully ionized gas '' which extends from the galactic center to 13kpc and 100pc above and below the galactic plane
. however continuum radiation from this particular medium was first discovered by westerhout(1958 ) in his dwingeloo survey , who also calculated the corresponding upper limits on density and total mass of ionized hydrogen in the galaxy .
later investigators have chosen to call this component as extended low density warm ionized medium(petuchowski & bennett 1993 , heiles 1994 ) .
the recent estimated densities of eldwim are @xmath5(murray & rahaman 2010 ) with temperatures of the order of 3000 - 8000k .
the origin , morphology and ionization mechanism of eldwim are uncertain .
eldwim has been thought to be as a collection of evolved hii regions(shaver 1976 ) or perhaps it forms the outer envelopes of hii regions(anantharamaiah 1985 ) . in the later argument
it has been suggested that by the sizes of hii region envelopes inferred from observations and due to their large number almost every line of sight intersects these envelopes in the inner galaxy ( @xmath7 ) giving the observed galactic ridge recombination lines(anantharamaiah 1986 ) .
more recently(murray & rahaman 2010 ) eldwim has been considered to be a diffuse gas ionized by massive stellar clusters unrelated to hii regions .
the present observations aimed at detecting helium lines from eldwim to understand its ionization spectrum .
the helium ionization potential(24.6 ev ) is higher than that of hydrogen(13.6 ev ) .
the detection and amplitude of he line would indicate the ionization condition of eldwim .
earlier rl observations have indicated that the ratio of the number of helium to hydrogen ions(@xmath8 ) in eldwim is in the range 0 to 0.054(heiles et.al 1996b ) .
this ratio is smaller than the generally accepted cosmic abundance of helium , @xmath9 0.1(poppi et.al 2007 and references therein ) . indicating all helium in eldwim to be partially ionized .
according to order of magnitude calculations(heiles et.al 1996b ; see also domgoergen & mathis 1994 ) this observed ratio in eldwim requires the surface temperature of the ionizing star to be @xmath335000k , if a standard hii region condition were to be considered .
the current knowledge of initial mass function and the total galactic star formation rate make it difficult to realize such a cool spectrum together with the total galactic ionization requirement for the eldwim and hii regions .
this has been called as the _ ionization problem _ ( heiles 1996b ) .
however such calculations are not completely exhaustive and alternate ways to explain the ionization spectrum must be found .
helium rls are much weaker than hydrogen rls and hence longer integration time is required to detect them from specific directions .
wsrt was used to observe hydrogen and helium rls from 15 different galactic positions . in the incoherent addition mode wsrt
offers 8 if bands , each with a bandwidth of 5 mhz .
these were used to detect 7 hn@xmath10/hen@xmath10 rls with n=165 to 171 and 1 hn@xmath11 rl with n=208 ( table-1 ) .
the primary objective was to detect the hen@xmath10 line by averaging the spectra from the 7 bands .
observations were carried out using dual frequency switching with a shift of 2mhz , keeping the average of the hydrogen and helium rest frequencies at the center of the band .
the resolution of the spectra is @xmath12 4km / s .
the 15 positions were constructed from a list of previously observed(lockman et.al 1976 , 1989 ; heiles 1996a ) directions which exhibited strong(@xmath050mk ) hydrogen rls with a single component at 1.4 ghz . a typical line strength of 50mk for hydrogen would imply @xmath13 = 0.025@xmath14 = 1.25mk for helium , taking into account the mean 0.025 of the earlier mentioned @xmath8 ratio .
with this strength for he lines the proposed integration time aimed at a 3@xmath2 detection .
care was also taken to avoid occurence of hii regions along the line of sight .
this was to ensure line origin from eldwim .
a list of relevant transitions and the emerging rest frequencies for hydrogen and helium rls is given in table-1 .
+ [ transitions ] .list of 8 transitions and rest frequencies corresponding to each band . [ cols="^,^,^",options="header " , ]
this work presents the results of hydrogen and helium rl observations towards 11 different galactic positions using wsrt at @xmath12 1.4ghz in incoherent addition mode . out of 11
, 5 positions exhibit helium rl detection with snr @xmath14@xmath2 .
wsrt provided 8 if bands each of width 5 mhz which were used to observe 8 distinct rl lines(hn@xmath10/hen@xmath10 with n=165 - 171 and h208@xmath11 ) using frequency switching
. details of observation and data analysis have been given in sec 2 & sec 3 .
these observations aimed at detecting helium rls from eldwim . since the abundance ( 0.1 ) of helium is smaller than that of hydrogen it is difficult to detect helium rls compared to the corresponding hydrogen rls . in view of this
a set of 15 positions was constructed by consulting previous rl observations ( lockman et.al 1976 , 1989 ; heiles et.al 1996a ) and the 11 cm continuum map(reich et al 1990 ) . in this procedure a region near to a previous strong(@xmath12 50mk )
hydrogen rl detection at 1.4 ghz was considered .
next during positioning of the beam on the 11 cm map care was taken not to include any possible hii region within the beam .
line detections from 11 of these positions have been displayed in figure 1 & 2 .
the fitted gaussian parameters are tabulated in table-2 , along with the obtained ratio @xmath8 . which is the ratio of line amplitudes of helium and hydrogen .
a detailed discussion of line parameters and their implication has been given in sec 4 .
in general most of the regions seem to be fit to represent eldwim .
this investigation has produced a high value(0.082 ) of @xmath8 ratio which is close to the accepted abundance(0.1 ) of helium to hydrogen .
indicating the presence of diffuse ionized helium in eldwim .
further a weighting scheme(eqn ( 1 ) ) has been adopted to show correlation between line orginating eldwim and the surrounding hii regions(paladini et.al 2003 & anderson et.al 2011 ) . according to this scheme
a weight is assigned to each region depending on ( i)the number of surrounding hii regions within a certain radius from the cloud ( ii)their distance from the cloud and ( iii)the distance(nearest ) of the cloud from the solar system(figure 5 ) .
this correlation has been displayed by plotting @xmath8 against the obtained weights , @xmath15 $ ] .
the conclusion from this plot is that regions with high value of @xmath8 bear a high weight . indicating a correlation between hii regions and helium line detection .
the author thanks the staff of wsrt who have made these observations possible .
the westerbork synthesis radio telescope is operated by the astron(netherlands institute for radio astronomy ) with support from the netherlands foundation for scientific research(nwo ) .
+ the author thanks the refree for his comments and suggestions that significantly improved the presentation of this paper . | helium line observations towards 11 galactic positions using westerbork synthesis radio telescope(wsrt ) have been reported .
these observations were made towards nearby positions where already hydrogen lines were detected at sufficiently high intensity(@xmath050mk ) at 1.4 ghz .
this approach gave a fair chance for the detection of helium line as well , keeping in mind the relative abundance(10 % ) of helium with respect to hydrogen .
care was also taken to avoid the presence of hii regions along the line of sight so that the line emission originates from the extended diffuse low density ionized component , eldwim of the galaxy .
the observations have resulted in the detection of helium line towards 5 positions out of 11 with signal to noise ratio(snr ) @xmath1 4@xmath2 .
an attempt has been made to associate detection / non - detection of helium line to the presence of surrounding hii regions .
a weighting scheme that accounts for nearby(@xmath3 500pc ) hii regions , their distances and other factors produces favourable results .
it is seen from this weighting scheme that a higher weight favours the detection of helium line while lower weight is associated with non - detection .
the idea is to correlate ionization of eldwim with the surrounding hii regions .
[ firstpage ] |
, response to the emerging carbon emissions constrained world , the usage of renewable energy sources is increasing . the overall increase in penetration of renewable energy resources in the u.s .
is depicted in fig .
[ fig : renewablepercentage ] .
such growth of renewable energy resource is not limited to the u.s .
globally , installed global renewable electricity capacity has continued to increase and represented 28.5% of total electricity capacity in 2014 @xcite @xcite .
renewable energy sources are generally characterized as _ variable energy resources _ ( ver ) due to their variability and uncertainty @xcite .
while there has been efforts for better control of resources such as wind farms as doubly fed induction generators @xcite , their limited controllability and lack of predictability pose new challenges for the operation of the power system . price - responsive demand , or _
demand response _
( dr ) is a key mechanism to achieve system balancing .
one avenue is by modifying consumption patterns through economically exposing customers to time - varying pricing that reflect supply - demand balancing status .
a number of programs on dr have been implemented or proposed @xcite @xcite . while such price - responsive demand can potentially provide the key to system operability under high penetration of vers , the presumed benefits of dr programs substantially depend on how responsive demand actually is to price : the price elasticity of demand @xcite . though price elasticity of demand is critical for the effectiveness of dr programs , previous empirical works based on data - driven _
static _ analysis of demand suggests that even load labeled as price - responsive is fairly inelastic @xcite . on the other hand , viewing dr as a _ dynamical system _ , our previous work @xcite analyzing industrial and commercial loads of which is directly exposed to _ electric reliability council of texas _ ( ercot ) real - time wholesale market makes two observations : 1 . the consumer s response to large prices ( over the 95%-quantile : $ 144.42 )
can be modeled as a _ hammerstein system _ , i.e. , a static nonlinearity followed by a linear transfer function @xcite . after accounting for this nonlinear transformation , which is typically concave since
the response is sublinear , the response exhibits a reduction after a delay of about 0.75 - 2.5 hours , before subsequently reverting back to normal levels .
the response to moderate prices ( up to $ 144.42 ) can be modeled as a linear stochastic system , specifically as an _ autoregressive exogenous _ ( arx ) system , i.e. , an autoregressive ( ar ) system with exogenous input and white noise .
this paper extends our previous work @xcite , by analyzing the economic effect of characteristics of consumer behavior that prevent real - time retail electricity pricing from optimal signaling and respon .
the rest of this paper is organized as follows . in section [ sec : litreview ] , previous works analyzing the models and benefits of price responsive demand , mostly conducted in the economics literature , are reviewed .
we introduce our observations on consumer behavior from empirical load data from ercot in section [ sec : consmobs ] . in section [ sec : effrtp ] , on the basis of our empirical observation we discuss the implication of our observations , presenting an alternative analysis of the potential benefits of dr in comparison to previous literature . concluding remarks followed in section [ sec : conc ] .
while the idea of dr is currently attracting wide interest as a solution for system operability under high penetration of vers , the necessity of dr has been advocated for decades by economists from a market efficiency perspective .
the volatility of load that has been challenge for system operators to cope with also entails abrupt and drastic changes in electricity price in the wholesale market . though extreme price fluctuation is widely observed in today s restructured electricity wholesale competitive markets ,
retail customers in most regions do not face frequent price change .
while wholesale electricity prices vary from hour to hour , retail prices do not change for months in most electricity markets .
such discordance between rapid fluctuation in wholesale prices and near flat retail prices not only incurs economic inefficiency in terms of social welfare , but also creates price - inelastic wholesale demand that severely exacerbates the volatility of wholesale electricity prices .
the combination of inelastic demand with the inherent real - time nature of the market makes electricity markets vulnerable to the exercise of market power @xcite . as a method to achieve price responsive demand ,
there has been a consensus on the potential benefits of _ real - time retail pricing _
( rtrp ) among economists @xcite @xcite @xcite @xcite @xcite @xcite .
the first potential benefit most discussed in the literature is the allocative efficiency improvement resulting from resolving the market inefficiency caused by ( near ) constant retail electricity prices @xcite @xcite @xcite @xcite @xcite @xcite .
the second benefit studied is the increased robustness of the market with rtrp forestalling the exercise of market power @xcite @xcite @xcite @xcite .
the last benefit considered is that the mitigation of demand volatility induced by real - time price signals will also relieve the need for excessive reserve requirement which incurs a large portion of the societal costs @xcite @xcite @xcite .
however , all the potential economic benefits of rtrp substantially depend on how responsive demand is to price , i.e. , the price elasticity of demand @xcite @xcite .
the efficiency improvement of rtrp is well analyzed in the literature @xcite @xcite @xcite @xcite , as depicted in figure [ fig : econinefffixrate ] . since the demand curve has a time variant property , it is not likely to happen that the fixed rate meets @xmath0 or @xmath1 , which are the optimal market clearing prices in terms of social welfare maximization . thus , the shaded triangles @xmath2 and @xmath3 are the deadweight loss , the economic inefficiency caused by the fixed rate @xmath4 . due to the instantaneousness of electricity
, it is reasonable to assume that the electricity at each time slot is a distinct commodity .
thus , rtrp advocates argue that the ultimate real - time retail price is the optimal pricing policy@xcite in terms of economic efficiency .
although the analysis shown in figure [ fig : econinefffixrate ] seems reasonable , it requires a crucial assumption to be justified : _ demand converges to @xmath0 or @xmath1 almost immediately , in at most one time slot as determined by the market rules .
_ however , this assumption is controversial when the market is fast - paced .
additionally , a fundamental limitation in the demand - supply curve model is that it is difficult to obtain any insight concerning dynamic behavior from the demand curve , which makes it difficult to estimate and predict demand from this static model .
the goal of our work is to develop a model for demand response as a stochastic dynamical system where both past prices and past consumption influence future consumption probabilistically .
in this section , we introduce our prior work @xcite that poses the problem of modeling price responsive demand at wholesale level . this work is based on an analysis of the data from an anonymous commercial / industrial ( c / i ) load in houston , purchasing its power directly from ercot real - time wholesale market , gleaned over nine months ( jan.1 - sep .
30 , 2008 ) . based on the empirical data , a dynamical model of consumer behavior is presented .
the c / i load and prices from houston measured at intervals of 15 minutes from jan . 1 , 2008 to sep .
30 , 2008 is presented with respect to time in fig .
[ fig : pq - hourlyplot ] .
the first notable point observed here is that the plot on price ( fig .
[ fig : price - boxplothourly ] ) shows many outliers while the plot on load rarely has any .
this is called the spiky " nature of electricity prices , an irregular sudden extreme price change for a very short duration of 15 - 30 minutes ( fig .
[ fig : price - sample ] ) .
this gives the price a highly non - normal heavy - tail distribution .
the fundamental reason for the spiky nature of prices is explained in section [ sec : litreview ] .
the other property we see in fig .
[ fig : pq - hourlyplot ] is that the time - series of load shows a depressed demand in `` peak hours '' ( afternoons ) , over time intervals that overlap with the time intervals exhibiting frequent large outliers in the price time series . here
, we surmise that the depressed demand is a manifestation of demand response , and that this demand response is highly connected to the outliers of price , because the depression is not likely to be explained by the plot of the median prices ( fig .
[ fig : price - medianhourly ] ) .
.statistics of price ( p ) and load ( @xmath5 ) [ cols="^,^,^,^,^ " , ] on the basis of the above observations , we establish a simple dynamic model between the magnitude of the price surge and the load , in the case of high price surges . taking into account the long - tailed characteristic of prices , we consider a linear model for a concave transformation @xmath6 , instead of @xmath7 . in this paper
, we present a tf for a specific time period , from 2:00pm to 2:30pm , due to the innate time - dependency on dr .
the estimation results for the arx model of dr at high price are presented in tables [ tab : dr - peak1 ] , [ tab : dr - peak2 ] , and [ tab : dr - peak3 ] .
the estimated tf of the arx model is : @xmath8 which accounts for 51.2% of the variance of @xmath9 .
the notable feature we find here is that the accuracy of the ar model for @xmath9 is severely degraded ( @xmath10 ) in table [ tab : dr - peak1 ] , compared to the ar model for the moderate price regime ( table [ tab : dr - moderate1 ] ) .
however , we observe that a relatively high portion ( 27% ) of the variance of @xmath11 is explained by the estimated model of @xmath11 shown in table [ tab : dr - peak2 ] , from which we conclude that the innovation from the price information is significant to improve @xmath12 of the arx model up to 51.2% , as presented in table [ tab : dr - peak3 ] .
ccccc + * coeff . &
* estimate & * se & * tstat & * pvalue + @xmath13 & 748.26 & 233.72 & 3.2015 & 0.0025097 + @xmath14 & 0.40153 & 0.11763 & 3.4133 & 0.0013678 + @xmath15 & -0.23826 & 0.1461 & -1.6308 & 0.10992 + @xmath16 & 0.25124 & 0.11516 & 2.1816 & 0.0344 + & + & + * * * * * ccccc + * coeff . &
* estimate & * se & * tstat & * pvalue + @xmath17 & 1213.4 & 293.68 & 4.1316 & 0.00014688 + @xmath18 & -220.1 & 52.774 & -4.1707 & 0.00012965 + & + & + * * * * * cccc + * coeff . & * estimate & * coeff . & * estimate + @xmath14 & 0.40153 & @xmath18 & -220.1 + @xmath15 & -0.23826 & @xmath19 & 1961.66 + @xmath16 & 0.25124 + & + * * * * in fig .
[ fig : dr - prediction ] , the validity of our model is shown by sample load forecast .
[ fig : dr - error - probplot ] and [ fig : dr - qqhat ] delineate the errors in the load forecast at 3:15pm after a price surge at 2:15pm .
we see that the forecasted @xmath20 and the actual @xmath9 at @xmath21 are reasonably well correlated ( correlation ( @xmath22 ) in fig .
[ fig : dr - qqhat ] , and that the errors exhibit normality ( kurtosis = 3.1809 ) in fig .
[ fig : dr - error - probplot ] .
our empirical study suggests that ( 1 ) _ the demand only responds to high price surges at peak hour _ and ( 2 ) _ there exists a demand response delay consequent on a high price surge . _
the second finding shows that there exists a certain `` inertia '' in consumption , resulting in a certain time delay to reduce power consumption after a peak price observation .
in this section , we shall further analyze the results from the data .
we will first examine the rationality of consumer behavior .
then we will show that the observed delay in demand response changes classical arguments about the role of prices and the equilibrium process , as well as classical efficiency results of markets .
one of our observations in section [ sec : consmobs ] is that price responsive demand exhibits delayed response to price shock at peak hours .
it may seem to be irrational to decrease one s demand after a price shock has already occurred .
however , if we consider that consumer behavior is based on _ prediction _ of price , rather than the current price itself , then we can explain the delayed response based on the inertia of demand . in this sense , decreasing one s demand after a price spike , specifically , during or after the price plummets after price surge ,
can be well explained as a rational behavior if there is a high chance of a price increase after price spike .
the chance of such a price increase relapse after a price spike is presented in figure [ fig : spikeprob ] .
figure [ fig : spikeprob ] shows a comparison of the estimated conditional probability of a high price in different situations , based on the obtained data from houston .
we can easily discern from figure [ fig : spikeprob ] that the conditional probability of a second price spike following the occurrence of a price spike quickly reduces in off - peak hours .
however , we also observe that the conditional probability of a price spike after the occurrence of a price spike during peak - hours remains at a significantly higher level than the probability of price spike without any conditioning .
this rationalizes our observation that _ if we assume a consumer has limited ability for immediate load reduction _ , then a rational consumer adjusts its load in response to price spike in spite of its inertia , because the relative chance of repeated price surge after a price spike is significantly high , and the demand is still not able to respond quickly to that following price surge .
this also explains why demand is responsive to price only when it is during peak hours .
the rigorous analysis of the consumer rationality behind such behavior will be presented in subsequent work .
based on the assumed rationality of a consumer , the premise that demand is an optimal choice in consumption space given a price bundle under individual s budget constraint , is the major basis for classical arguments regarding the market benefits of price responsive demand . from the viewpoint of classical arguments , the demand curve as a function of price always represents an optimal solution that maximizes ( aggregated ) consumer utility given a budget constraint .
however , the existence of inertia of demand suggests that such inertia may result in a restriction of feasible choices in consumption space , which then prevents consumer from engaging in the optimal consumption choice predicted by classical consumer theory .
this implies that an instantly observed static demand curve may not fully reflect consumers utility structure regardless of consumers rationality , which may impact negatively on overall market efficiency under previously suggested pricing solutions for electricity markets . while there may exist a _
nominal demand curve _
well reflecting a consumer s utility , the inability to respond instantly creates a temporal distortion in demand curve in two respects : ( 1 ) since the instantaneous price elasticity of demand is near zero , an instantaneous demand curve during the time slot with a sudden price surge may be represented by a vertical line in a classical quantity - price plot , so that it may not coincide with the nominal demand curve ; and ( 2 ) the demand hedged against the risk of a repeated price surge would not respond to subsequent price reduction , so that an instantaneous demand curve after an unruffled price may not coincide with the nominal demand curve for a certain subsequent period of time .
it is generally expected that dr will be beneficial both in terms of system operability as well as economical perspective , under the high penetration of vers .
however , the distortion in the demand curve due to demand inertia can degrade the overall expected benefits from dr programs designed incautiously without consideration of demand inertia .
an important example of the latter is high - frequency rtrp . in this subsection
, we discuss the limitation of potential benefits from high - frequency rtrp under supply fluctuation caused by vers .
while the analysis on the allocative efficiency of rtrp compared to fixed price under volatile demand analyzed in previous literature , is depicted in figure [ fig : econinefffixrate ] , this analysis can be extended to the allocative efficiency of rtrp compared to fixed price under supply fluctuation by vers as depicted in figure [ fig : econinefffixratever ] .
analogously , the shaded triangles @xmath2 and @xmath23 are the deadweight loss , representing the economic inefficiency caused by the fixed rate @xmath4 , or , conversely , the expected allocative efficiency benefit from rtrp compared to a fixed tariff . according to the classical arguments on rtrp , more frequent price change would be more beneficial , because they would more accurately the balance supply and demand in real time , so that it is more informative for consumers to make an optimal decision .
however , our observation suggests that such allocative efficiency is not likely to be achievable because of the distortion in demand curve caused by demand inertia .
the inability of customers to respond instantly may distort the demand curve to a vertical line in quantity - price plot . noting this distortion , the demand behavior in practice
may be realized as if it is exposed to fixed price .
the impact on allocative efficiency from the demand curve distortion is presented in figure [ fig : ineffver ] .
figure [ fig : ineffverdropoff ] depicts the situation where ver drops due to events such as sudden diminution of wind or cloud cover blocking the sun .
suppose that the market equilibrium point before the ver drop is @xmath24 .
then , the ver drop shifts the supply curve to the left . after the supply curve shift ,
the optimal market clearing point maximizing social welfare is @xmath0 on the nominal demand curve .
however , the actual market clearing point is realized at @xmath25 due to demand curve distortion .
hence , the shaded triangle @xmath2 is the deadweight loss , exhibiting the economic inefficiency following from demand inertia . in figure
[ fig : ineffverrestoration ] , the analysis of the situation where ver is restored by incidents e.g. the increase in wind generation caused by a gust of wind , or that of solar generation following a cloud gap is provided
. suppose the market equilibrium point before ver restoration is @xmath24 .
the increase in generation followed by ver restoration event results in the newly formed supply curve to lie on the right hand side of the previous one .
while @xmath0 on the nominal demand curve is the optimal market clearing point after the supply curve shift , the demand curve distortion caused by the hedging of demand against the risk of a repeated ver drop - off may result in the actual market clearing point being realized at @xmath25 .
again , the shaded triangles @xmath2 indicates the deadweight loss , the economic inefficiency from demand inertia .
the allocative ( in)efficiency analysis of rtrp under demand inertia suggests that there is a fundamental limitation to achieving market efficiency that can be expected from traditional market efficiency analysis without consideration of demand inertia .
this necessitates a redefinition of market efficiency from an optimal control theoretical perspective .
in addition , the demand behavior under rtrp is as if it is exposed to a fixed price , leading to another crucial implication that rtrp may not significantly resolve the vulnerability of markets from the exercise of market power .
the differences in the ability of various market participants to control their behavior endows different market powers to each market participant ; the more instantaneously responsive market participant has an advantage over market participants with larger delay .
such a combination of differentially endowed market powers makes market more vulnerable to the exercise of market power or market manipulation . a similar argument is found in financial markets with high - frequency trading ( hft ) practices , in terms of the robustness with respect to market manipulation and market fairness @xcite .
moreover , the inability of demand response to instantaneously respond also suggests that rtrp may contribute negatively to demand volatility mitigation , so that the savings in the cost of maintaining reserve capacity may be less than expected under previous literature .
a market is a dynamical system that is designed to proceed toward an optimal state as its equilibrium .
however , such a process necessarily requires a certain amount of time to reach its equilibrium . while dynamic modeling and control on the generation side in power systems has been well understood , the understanding of dynamic behavior on the demand side in response to price has been unclear . in this paper
, we consider a consumer s dynamic behavior in response to real time price change , by studying empirical data on a price - responsive load in the ercot area . out
empirical study suggests the following : ( 1 ) _ the price responsiveness of demand may exhibit qualitatively different behavior in response to `` normal price '' and `` high price '' ; _ and ( 2 ) _ there exists a demand response delay consequent on a high price surge at peak hours . _ this idea provides important guidance for designing two fundamental factors in time - varying retail electricity prices , _ frequency _ and _ timeliness_. here frequency of price " is the frequency at which retail prices change , and timeliness of price " is the time lag between when a price is set and when it is effective @xcite .
it is generally assumed among economists that rtrp with high frequency and just - in - time timeliness would be ideal in terms of economic efficiency in the electricity market , as rtrp is an attempt to provide more accurate signals closely reflecting the actual supply / demand status in the market . however , the inference based on our work is that neither argument is necessarily right .
the inherent delay in the responsiveness of loads to high price volatility exacerbates the predictability of price , thereby making demand less responsive to rtrp , which in fact worsens economic efficiency .
consumers which are more exposed to market volatility stiffen their demand to be more inelastic and tend to be more conservative due to the inertial nature of demand .
this suggests that there exists a trade - off between controllability of demand and observability of markets , so that there may exist an optimal frequency and timeliness which should be carefully considered for optimal pricing design .
this also supports the importance of relatively long - term contract markets such as day - ahead electricity markets .
market efficiency should be re - analyzed taking into consideration the trade - off between the controllability of demand and the observability of the market . in subsequent work
, we aim to provide a rigorous analysis of consumer rationality and develop a quantitative prediction model for demand response .
frank a. wolak , `` managing demand - side economic and political constraints on electricity industry restructuring processes , '' _ economic reform in india , cambridge university press _ , cambridge , uk , ch .
455 - 498 , 2013 . c. aubin ,
d. fougere , e. husson , and m. ivaldi , `` real - time pricing of electricity for residential customers : econometric analysis of an experiment , '' _ journal of applied econometrics _
171191 . dec . | in this paper , we study the price responsiveness of electricity consumption from empirical commercial and industrial load data obtained from texas . employing a dynamical system perspective
, we show that price responsive demand can be modeled as a hybrid of a hammerstein model with delay following a price surge , and a linear arx model under moderate price changes .
it is observed that electricity consumption therefore has unique characteristics including ( 1 ) qualitatively distinct response between moderate and extremely high prices ; and ( 2 ) a time delay associated with the response to high prices .
it is shown that these observed features may render traditional approaches to demand response and retail pricing based on classical economic theories ineffective .
in particular , ultimate real - time retail pricing may be limitedly beneficial than as considered in classical economic theories .
demand response , electricity market , dynamic system modeling . |
double parton collisions are a new feature of high energy hadron interactions which becomes increasingly important at high energies .
the number of partons that can undergo a hard collision is in fact a fast growing function of @xmath1 and , as a consequence , when the c.m .
energy is large enough , the probability of having more than one hard partonic interaction in the same inelastic event becomes sizable .
double parton collisions , foreseen long ago by several authors@xcite , have been in fact observed recently by cdf@xcite .
the non - perturbative input to the double parton collisions are the double parton distributions , which are independent on the parton distributions usually considered in large@xmath2 physics , since they are related directly to the two - body parton correlation of the hadron structure . in the simplified hypothesis of neglecting all correlations in fractional momenta , the inclusive double parton scattering cross section for the two parton processes @xmath3 and @xmath4 reduces nevertheless to a simplest factorized expression : @xmath5 where @xmath6 when @xmath3 and @xmath4 are indistinguishable processes and @xmath7 when they are distinguishable . @xmath8 and @xmath9 are the inclusive single scattering cross sections for producing the processes @xmath3 and @xmath4 respectively and all the new information on the structure of the hadron in transverse space is summarized in the value of the scale factor @xmath0 .
this simplest hypothesis has not been contradicted by the experimental evidence@xcite , whose results are in fact described with a single parameter ( the scale factor @xmath0 ) .
the experimental value quoted by cdf , @xmath10 mb , is however too small to be understood in a simplest uncorrelated picture of the multi - parton distribution and indicates that correlations in transverse space play an important role@xcite .
a possibility which has been considered@xcite to explain the smallness of the value of @xmath0 is to correlate the population of gluons and sea quarks with the configuration of the valence in transverse space , in such a way that the average number of gluons and sea quarks is small when the valence quarks are all close in transverse space and , on the contrary , is large when they are separated by a ( relatively ) large transverse distance .
such a mechanism increases the dispersion in the number of multiple parton collisions and , as a consequence , @xmath11 ( which is proportional to that dispersion@xcite ) .
the value of @xmath0 is therefore diminished with respect to the uncorrelated case .
the scale factor , on the other hand , is the result of the overlap in transverse space of the matter distribution of the two interacting hadrons and a feature of the model above is that the average transverse distance of a pair of valence quarks is different as compared with the average transverse distance of a pair of gluons or sea quarks . in the model @xmath0 is therefore different for double parton scatterings involving valence quarks and for double parton scattering involving sea quarks and gluons . although the details of the model should be understood in a qualitative rather than in a quantitative sense , it is rather natural to expect different values of the scale factor in different reactions .
we think therefore that it may be interesting to have an indication , even if only qualitative , on the size of the effect and , to that purpose , we work out in the present note the values of the scale factor foreseen in the model @xcite in a few cases of interest .
at large energies one will be able to observe double parton collisions in various channels@xcite@xcite@xcite . in particular the production of two equal sign @xmath12 bosons , at relatively low @xmath13 , goes almost entirely through double parton collisions@xcite and is originated , to a large extent , by interactions with valence quarks . another case
where double parton collisions are expected to play an important role is in the production of two @xmath14 pairs .
the corresponding integrated cross section is in fact rather large , by using eq.[fact ] one may estimate that the double parton scattering cross section for producing two @xmath14 pairs is of the order of @xmath15 at the energy of the lhc .
the dominant mechanism of @xmath16 production is gluon fusion and the observation of double parton collisions , in a equal sign @xmath12 boson pair production and in the production of two @xmath14 pairs , allows therefore one to compare the distribution in transverse space of valence quarks with the distribution in transverse space of gluons . to work out the scale factor in the two cases
we write the inclusive double parton scattering cross section , for the parton interactions @xmath3 and @xmath4 , as @xmath17 where @xmath18 are the double parton distributions , depending on the two fractional momenta of the partons @xmath19 and @xmath20 and on their relative transverse distance @xmath21 .
the indices @xmath22 and @xmath23 refer to the different kinds of partons , @xmath24 and @xmath25 are the partonic cross sections and the qcd dependence on the scale of the interaction is implicit in all quantities . in the case we are considering
the dependence of @xmath26 on @xmath19 , @xmath20 and @xmath21 is factorized : @xmath27 where @xmath28 are the usual parton distributions .
if @xmath29 do not depend on @xmath22 and @xmath23 one obtains eq.[fact ] and the scale factor @xmath0 is universal . in general
the double scattering cross section is however written as @xmath30 where @xmath31 is the hadronic inclusive cross section for two partons of kind @xmath22 and @xmath23 to undergo the hard interaction @xmath3 , and @xmath32 are geometrical coefficients , with dimensions of the inverse of a cross section , depending on the various parton processes . in the model
@xcite the probability to find the proton in a given configuration , with @xmath33 gluons or sea - quarks , has the following expression : @xmath34 { \rm exp}\bigg\{-\frac{b_d^2}{\langle b_d^2\rangle}\int g(x)dx \bigg\ } \label{pdistr2}\end{aligned}\ ] ] where @xmath35 are the inclusive distributions of valence quarks , as a function of the momentum fraction @xmath36 , and @xmath37 is the density of valence quarks in transverse space , where the coordinates of valence quarks in transverse space @xmath38 are given by @xmath39 in the model @xmath40 , where @xmath41 represents the distribution density of the proton in coordinate space .
the explicit expression used is @xmath42 where @xmath43 with @xmath44fm , the proton charge radius . given a configuration of the valence , the average density of gluons and sea quarks in a point with transverse coordinate @xmath21 and momentum fraction @xmath45 is given by @xmath46 .
the overall average number of gluons and sea quarks at a given @xmath45 ( namely after integrating on @xmath21 and on the configurations of the valence ) is @xmath47 and is identified with the inclusive distribution of gluons and sea quarks . in the same way the inclusive distributions of the valence quarks , @xmath35 , are the result of integrating over
the transverse coordinates @xmath38 and of summing over all configurations of gluons and sea quarks .
the dependence of the average density of gluons and sea quarks on the transverse coordinate @xmath21 is expressed by @xmath48 which is the projection on the transverse plane of a sphere of radius @xmath49 , rescaled with the factor @xmath50 , where the average @xmath51 is taken with the density @xmath37 .
the average density of gluons and sea quarks and the corresponding average number ( which grows as @xmath50 ) are therefore correlated with the configuration of the valence , while all fractional momenta are uncorrelated .
the model distinguishes two different kinds of partons , as far as the number and density in transverse space are concerned , the valence quarks and the sea quark and gluons . the indices @xmath22 and @xmath23 , as a consequence , can take two different values , @xmath52 and @xmath1 , and the relevant transverse densities of parton pairs to be considered are @xmath53 by using eq.[fb ] and the expression below , the matrix @xmath54 and the effective cross section are readily evaluated : @xmath55
the resulting values of @xmath56 are shown in the table .
the scale factor @xmath0 is plotted as a function of the c.m .
energy in fig.1 and in fig.2 , corresponding to @xmath57 and @xmath58 interactions respectively , in various processes of interest : @xmath22 ) production of two equal sign @xmath12 bosons , @xmath59 ) production of a @xmath12 boson of either positive or negative sign together with two jets or @xmath60 ) with a @xmath16 pair , @xmath61 ) production of four jets , @xmath52 ) production of two @xmath16 pairs . in the case of jets
the scale factor has been evaluated by using as a lower cutoff in @xmath13 the value of @xmath62 ( the scale factor is however rather insensitive to that choice ) .
all the single scattering cross sections in eq.[seff ] have been computed at the lowest order in the coupling constant and by making use of the martin - roberts - stirling 1999 ( mrs99 ) parton distributions@xcite .
the different results obtained for @xmath63 and @xmath64 production for @xmath57 collisions is mainly due to the different content of @xmath65 and @xmath66 quarks in the proton .
a consequence is in fact the different contribution of the sea quarks in the two cases , whose distribution is sizably different in the model in comparison with the distribution in transverse space of the valence . to check the dependence on the choice of the distribution functions we have repeated the calculation by using the cteq5 parton distributions@xcite . in all cases the scale factor
is not changed by more than a few percent .
in fact the results for @xmath0 are rather insensitive to the choice of parton distributions , since @xmath0 is obtained by making ration of cross sections .
the main qualitative feature of the results obtained is the strong difference , at tevatron energy , between final states with and without a @xmath12 boson , and the energy dependence of the scale factor , which is sizable when moving from tevatron to lhc energy in the channels containing a @xmath12 boson . while the actual values should be considered only as indicative , the qualitative features just pointed out are likely to be of more general validity , since they are originated by the different source of initial state partons in the case of @xmath12 production , which in most cases involves valence quarks , in comparison with the other channels considered , that are mainly generated by interactions of sea quarks and gluons .
the overall indication is that it is plausible to expect non minor differences in the values of @xmath0 in different processes .
it is also apparent that a basic element , to understand the three dimensional structure of the proton , is a quantitative information on the correlations in transverse space of the various pairs of initial state partons and that the scale factors , of the different double parton collision processes , are the physical observables which can provide that information .
landshoff and j.c .
polkinghorne,_phys . rev . _
* d18 * 3344 ( 1978 ) ; fujio takagi _ phys . rev
. lett . _ * 43 * , 1296 ( 1979 ) ; c. goebel , f. halzen and d.m .
scott , _ phys .
rev . _ * d22 * , 2789 ( 1980 ) ; n. paver and d. treleani , _
nuovo cimento _ * a70 * , 215 ( 1982 ) ; b. humpert , _ phys .
lett . _ * b131 * , 461 ( 1983 ) ; m. mekhfi , _ phys .
rev . _ * d32 * , 2371 ( 1985 ) , _ ibid .
_ * d32 * , 2380 ( 1985 ) ; b. humpert and r. odorico , _ phys .
_ * 154b * , 211 ( 1985 ) ; t. sjostrand and m. van zijl , _ phys .
* d36 * , 2019 ( 1987 ) ; f. halzen , p. hoyer and w.j .
_ * 188b * , 375 ( 1987 ) ; m. mangano , _ z. phys . _ * c42 * , 331 ( 1989 ) ; r.m .
godbole , sourendu gupta and j. lindfors , _ z. phys . _ * c47 * 69 ( 1990 ) . f. abe et al .
( cdf collaboration ) , _ phys.rev.lett . _ * 79 * 584 ( 1997 ) ; _ phys .
* d56 * 3811 ( 1997 ) . g. calucci and d. treleani , phys . rev .
* d57 * 503 ( 1998 ) . g. calucci and d. treleani , phys .
* d60 * ( 1999 ) 054023 .
s. catani , m. dittmar , d. soper , w.j .
stirling , s. tapprogge , s. alekhin , p. aurenche , c. balazs , r.d .
ball , g. battistoni , e.l .
berger , t. binoth , r. brock , d. casey , g. corcella , v. del duca , a. del fabbro , a. de roeck , c. ewerz , d. de florian , m. fontannaz , s. frixione , w.t .
giele , m. grazzini , j.p .
guillet , g. heinrich , j. huston , j. kalk , a.l .
kataev , k. kato , s. keller , m. klasen , d.a .
kosower , a. kulesza , z. kunszt , a. kupco , v.a .
ilyin , l. magnea , m.l .
mangano , a.d .
martin , k. mazumdar , ph .
mine , m. moretti , w.l .
van neerven , g. parente , d. perret - gallix , e. pilon , a.e .
pukhov , i. puljak , j. pumplin , e. richter - was , r.g .
roberts , g.p .
salam , m.h .
seymour , n. skachkov , a.v .
sidorov , h. stenzel , d. stump , r.s .
thorne , d. treleani , w.k .
tung , a. vogt , b.r .
webber , m. werlen , s. zmouchko .
cern - th-2000 - 131 , hep - ph/0005025 .
a. del fabbro and d. treleani , phys .
rev . * d61 * ( 2000 ) 077502 .
a. kulesza and w.james stirling , phys .
* b475 * ( 2000 ) 168 .
a.d.martin , r.g.roberts , w.j.stirling and r.s.thorne , eur .
j. * c14 * ( 2000 ) 133 .
h. l. lai , j. huston , s. kuhlmann , j. morfin , f. olness , j. f. owens , j. pumplin , w. k. tung , eur .
j. * c12 * ( 2000 ) 375 . | the scale factor @xmath0 , which characterizes double parton collisions in high energy hadron interactions , is a direct manifestation of the distribution of the interacting partons in transverse space , in such a way that different distributions give rise to different values of @xmath0 in different double parton collision processes .
we work out the value of the scale factor in a few reactions of interest , in a correlated model of the multi - parton density of the proton recently proposed .
e - mail [email protected] + e - mail [email protected] + |
the fact that we live in an accelerating expanding universe is a well established reality today .
the direct evidence of such an accelerating expansion of the universe first published by the high - z supernova search teams ( riess et al .
1998 ; perlmutter et al . 1999 ) in 1998 and early 1999 .
later on , some other observations such as measurements of cosmic microwave background ( miller et al . 1999 ; bernardis et al . 2000 ; halverson et al .
2002 ; benoit et al . 2003 ; mason et al . 2003 ) and the galaxy power spectrum ( spergal et al . 2003 ; tegmark et al . 2004 ; page et al . 2003 ) indicate that the expansion of our universe is speeding up currently .
these observations strongly suggest that we live in a nearly spatially flat universe with approximately @xmath7 of the critical energy density being in non - relativistic matter and @xmath8 in a smooth component with high negative pressure called dark energy ( de ) .
the concept of dark energy refers to a kind of exotic energy with negative pressure which has been proposed for understanding the current accelerating expansion of the universe .
although , dark energy has very significant effect on the evolution of the universe at large scales and the ultimate fate of our universe will be determined by this unknown component , our knowledge about its nature and properties is still very limited .
we only know the following well established information about dark energy ( i ) it is an exotic energy with negative pressure which is caused the present cosmic acceleration ; ( ii ) it is non - clustering and spatially homogeneous ; and ( iii ) it was small at the early times while dominates the present universe ( at @xmath9 ) .
+ the study of dark energy is possible through its equation of state ( eos ) parameter @xmath10 . here , @xmath11 and @xmath12 are the pressure and the energy density of de respectively . unfortunately , the current value of the dark energy eos parameter is not known to us yet .
this is why so far many candidates have been proposed for dark energy .
the first candidate for dark energy which can simply explain the observed cosmic acceleration is a cosmological constant ( weinberg 1989 ; carroll 2001 ; padmanabhan 2003 ; peebles and ratra 2003 ) with @xmath13 .
however some fundamental theoretical problems , such as the fine - tuning problem and the coincidence problem associated with cosmological constant lead cosmologists to think about another candidates such as quintessence ( @xmath14 ) ( wetterich 1988 ; ratra and peebles 1988 ) , phantom ( @xmath15 ) ( caldwell 2002 ) , quintom ( @xmath16 ) ( feng et al .
2005 ) , chaplygin gas as well as generalized chaplygin gas models ( srivastava 2005 ; bertolami et al .
2004 ; bento et al .
2002 ; alam et al .
2003 ) , and etc .
quintessence dark energy models introduce a scalar field @xmath17 that is minimally coupled to gravity .
although , a possible solution to the cosmic coincidence problem my be provided by the attractor version of quintessence , but as dark energy , quintessence field with @xmath18 is not in agreement with the recent observations ( hinshaw et al .
2009 ; komatsu et al . 2009
; copeland et al . 2006 ; perivolaropoulos 2006 ) which indicate that @xmath15 is allowed at @xmath19 confidence level .
since the present observational data do not exclude the possibility of @xmath20 to be less than @xmath21 , a new class of scalar field models with negative kinetic energy , known as the phantom field models ( caldwell 2002 ) have been introduced .
this scenario also would encounter two essential problems namely the big rip ( caldwell et al .
2003 ; nesseris and perivolaropoulos 2004 ) of the universe and the ultraviolet quantum instabilities problem ( carroll et al .
since the models with a transition from @xmath22 to @xmath15 are mildly favored by recent observations , a new scenario of dark energy dubbed quintom was proposed ( feng et al . 2005 ) .
this model can be considered as a combination of quintessence and phantom dark energy models which can fulfill the transition from @xmath22 to @xmath15 and _ vice versa_. + recent astrophysical observational evidences such as the large entropy per baryon and the remarkable degree of the isotropy of cosmic microwave background radiation ( cmbr ) indicate that the cosmic media is not a perfect fluid ( jaffe et al .
these observed physical phenomena reveal the importance of analysis of the dissipative effects in cosmology .
it is believed that the occurrence of the big rip can be avoided in dissipative dark energy models in which the negative pressure is an effective bulk viscosity ( mcinnes 2002 ; barrow 2004 ) .
zimdahl et al .
( 2001 ) and balakin et al .
( 2003 ) have studied the role of viscous pressure as an agent that drives the present acceleration of the universe .
padmanabhan and chitre ( 1987 ) have also investigated the possibility of a viscosity dominated late epoch of the universe with accelerated expansion .
the effect of bulk viscosity on the background expansion of the universe has been studied from different points of view ( cataldo et al .
2005 ; brevik ans gorbunova 2005 ; szydlowski and hrycyna 2007 ; singh 2008 ; feng and zhou 2009 ; oliver et al . 2011 ; amirhashchi 2013 ) .
it is worth to mention that the general theory of dissipation in relativistic imperfect fluid was first suggested by eckart ( 1940 ) , landau and lifshitz ( 1987 ) .
+ it is believed that an energy transition from dark energy to dark matter could lead to a solution to the coincidence problem in cosmology ( cimento et al .
2003 ; dalal et al . 2001 ; jamil and rashid 2008 ; jamil and rashid 2009 ) . in other word , by considering a coupling between dark energy and dark matter , we can explain why the energy densities of de and dm are nearly equal today .
it is worth to mention that the recent observations ( bertolami et al .
2007 ; le delliou et al . 2007
; berger and shojaei 2006 ) provide the evidence for the possibility of such an interaction between these two dark components . from theoretical point of view
, the possibility of the interaction between de and dm have been widely investigated in the literatures .
zhang ( 2005a , 2005b ) , zimdahl and pavon ( 2004 ) , setare ( 2007a , 2007b , 2007c , 2009a , 2009b ) , setare et al .
( 2009 , 2011 ) , banijamali et al .
( 2009 ) , sheykhi and setare ( 2010 , 2011 ) , pradhan et al ( 2011a , 2011b ) , saha et al .
( 2012 ) , and amirhashchi et al .
( 2011a , 2011b , 2011c ) have investigated dark energy in different contexts .
viscous dark energy and generalized second law of thermodynamics as well as the thermodynamics of viscous dark energy have been recently studied by setare and shikhi ( 2010a , 2010b ) .
motivated the situation discussed above , in this paper , we study the impact of the interaction between viscous de and dm on the equation of state ( eos ) parameter of dark energy in the scope of anisotropic bianchi type i universe .
we consider an energy flow from de to dm and parameterize the interaction by a constant @xmath1 and viscosity by @xmath0 , then a detailed investigation of the cosmological implications of this parametrization will be provided . as usual , a statefinder diagnostic will be performed in order to discriminate the different interaction parameters .
finally , by using @xmath5 method we investigate the constraints on the model parameters @xmath1 and @xmath4 utilizing the recent observational data of hubble rate @xmath23 .
in an orthogonal form , the bianchi type i line - element is given by @xmath24 where @xmath25 and @xmath26 are functions of time only .
+ the einstein s field equations ( in gravitational units @xmath27 ) read as @xmath28 where @xmath29 and @xmath30 are the energy momentum tensors of dark matter and viscous dark energy , respectively .
these are given by @xmath31,\ ] ] @xmath32\rho^{m},\ ] ] and @xmath33,\ ] ] @xmath34\rho^{x},\ ] ] where @xmath35 and @xmath36 are the energy density and pressure of the perfect fluid component while @xmath37 is its eos parameter .
similarly , @xmath12 and @xmath11 are , respectively the energy density and pressure of the viscous de component while @xmath38 is the corresponding eos parameter . in this paper
we assume the following expression for the pressure of the viscous fluid ( eckart 1940 ) @xmath39 where @xmath40 is the viscous pressure and @xmath41 is the hubble s parameter . here
@xmath42 is the four velocity vector satisfying @xmath43 .
it is important to note that as a consequence of the positive sign of the entropy change in an irreversible process , @xmath44 has to be positive ( nojiri and odintsov 2003 ) . in general , @xmath45 , where @xmath46 and @xmath47 are constant parameters .
+ in a co - moving coordinate system ( @xmath48 ) , einstein s field equations ( [ eq2 ] ) with ( [ eq3 ] ) and ( [ eq4 ] ) for bianchi type - i metric ( [ eq1 ] ) subsequently lead to the following system of equations : @xmath49 @xmath50 @xmath51 @xmath52 solving eqs .
( [ eq6])-([eq9 ] ) , one can find ( saha 2005 ) @xmath53 @xmath54 and @xmath55 where @xmath56 here @xmath57 is the average scale factor of bianchi type i model .
using eqs .
( [ eq10])-([eq12 ] ) in eq .
( [ eq9 ] ) we obtain @xmath58 which is the analogue of the friedmann equation and @xmath59 , is a constant .
note that @xmath60 denotes the deviation from isotropy e.g. @xmath61 represents flat flrw universe .
+ now we assume the following model - independent deceleration parameter ( xu et al .
2006 ) @xmath62 where @xmath63 and @xmath64 are constants determined by the recent observational constraints . from ( [ eq14 ] )
we observe that for @xmath65 , @xmath66 which is corresponding to matter dominated era whereas for @xmath67 , the current value of deceleration parameter is obtained as @xmath68 . in view of eq .
( [ eq14 ] ) and @xmath69 the hubble parameter is obtained as @xmath70.\ ] ]
in this section we study the behavior of viscous dark energy when there is an energy flow from de to dark matter ( dm ) . in this case
the law of energy - conservation equation ( @xmath71 ) which yields @xmath72 can be written in the following form for barotropic and dark fluids @xmath73 @xmath74 separately . here the interaction between the dark components is shown by the quantity @xmath75 . to find a solution for the coincidence problem
, we have to consider an energy transfer from the dark energy to the dark matter by assuming @xmath76 .
@xmath76 , ensures that the second law of thermodynamics is fulfilled ( pavon and wang 2009 ) .
we note that from eqs .
( [ eq18 ] ) and ( [ eq19 ] ) it is easy to find that the unit of @xmath75 may be of inverse of time .
hence , a first and natural candidate can be the hubble factor @xmath6 multiplied with the energy density . in this paper , following amendola et al .
( 2007 ) and gou et al .
( 2007 ) , we consider and the dark energy density @xmath12 for @xmath77 , @xmath78 , @xmath79 , @xmath80 , @xmath81 , @xmath82 , and @xmath83.,title="fig : " ] + and the dark energy density @xmath12 for @xmath77 , @xmath78 , @xmath79 , @xmath80 , @xmath81 , @xmath82 , and @xmath83 . ]
@xmath84 where @xmath1 is a coupling coefficient which can be considered as a constant or a function of red shift .
the combination of the snls , cmb , and bao databases marginalized over a present dark energy density gives stringent constraints on the coupling , @xmath85 ( @xmath86 c.l . ) in the constant coupling model and @xmath87 ( @xmath86 c.l . ) in the varying coupling model , where @xmath1 is a present value ( guo et al .
+ we note that the eq .
( [ eq20 ] ) is not the only choose for @xmath75 as one can select the following forms of @xmath75 : ( i ) @xmath88 and ( ii ) @xmath89 .
the freedom of choosing @xmath75 is because of our lack of knowledge of the nature of de and dm .
+ using eq .
( [ eq20 ] ) in eq .
( [ eq18 ] ) and after integrating , we obtain @xmath90 where @xmath91 is an integrating constant .
+ by using eqs .
( [ eq21 ] ) and ( [ eq16 ] ) in eq .
( [ eq13 ] ) , we obtain @xmath92-\omega_{0}^{m } ( 1+z)^{3(1+\omega^{m}-\sigma)}\right)-3k(1+z)^6,\ ] ] where we have used @xmath93 and @xmath94 .
+ now , using eqs .
( [ eq21 ] ) and ( [ eq22 ] ) in eq .
( [ eq6 ] ) , the effective eos of dark energy @xmath95 is obtained as @xmath96 @xmath97-k(1+z)^{6}+\omega^{m}\rho^{m}}{3 h^{2}_{0}\left((1+z)^{3}\mbox{exp}\left[\frac{2\alpha((1+z)^{-\beta}-1)}{\beta}\right]-\omega_{0}^{m } ( 1+z)^{3(1+\omega^{m}-\sigma)}\right)-3k(1+z)^6}-\zeta_{0}h^{\tau}(\omega^{x})^{\tau-1}.\ ] ] versus @xmath98 for @xmath77 , @xmath78 , @xmath79 , @xmath80 , @xmath81 , and @xmath82 . in this case
, we fix @xmath99 and vary @xmath0.,title="fig : " ] + versus @xmath98 for @xmath77 , @xmath78 , @xmath79 , @xmath80 , @xmath81 , and @xmath82 . in this case
, we fix @xmath99 and vary @xmath0 . ]
note that here @xmath100 , @xmath101 , and @xmath102 .
+ figures @xmath103 show the phase - plane diagram representing the evolution of the dark matter density @xmath35 and the dark energy density @xmath12 . in fig .
1 we have assumed that there is no interaction between dark components i.e @xmath104 . in this case
the dark matter is continually being converted into the dark energy @xmath12 , hence as @xmath35 monotonically tends to @xmath105 , @xmath12 tends monotonically to @xmath106 .
therefore , the ratio @xmath107 always increases which aggravates the coincidence problem .
but as it is shown in fig .
2 , interaction between matter and dark energy could lead to the energy transfer from dark energy to dark matter .
the scenario of converting dark matter to dark energy and vis versa may occurs several times providing a cyclic decreasing - accelerating model of our universe .
such a model would provide a solution to the coincidence problem since the dark energy and dark matter will be comparable at infinitely several times .
+ the behavior of eos ( @xmath95 ) parameter for dark energy in terms of red shift @xmath98 is depicted in figures .
since we are interested in the late time evolution of de , we plot the range of red shift @xmath98 from @xmath109 to @xmath67 and the parameter @xmath110 is taken to be @xmath105 . in fig .
@xmath111 we fix the parameter @xmath112 and vary @xmath1 as @xmath105 , @xmath113 , @xmath114 , and @xmath115 respectively ; in fig .
@xmath116 we fix @xmath83 and vary @xmath0 as @xmath105 , @xmath117 , @xmath118 , and @xmath119 respectively .
the plots show that the evolution of @xmath95 depends on the parameters @xmath1 and @xmath0 apparently . in another word , from fig .
@xmath108 we observe that the effective eos parameter of an interacting viscous de model varies from quintessence region to phantom region and ultimately approaches to the same value of @xmath95 depending on the value of the bulk viscous coefficient @xmath0 .
we also observe that either in non - interacting case with small @xmath0 or interacting case with small @xmath1 , the eos parameter only varies in quintessence region .
it is worth to mention that while the current cosmological data from snia ( riess et al .
2004 ; astier et al . 2006 ) , cmb ( komatsu et al . 2009 ; rodrigues 2008 ) , and sdss ( cruz et al . 2007 ) rule out that @xmath120 , they mildly favor a dynamically evolving de crossing the pde
. therefore , our results are in good agreement with well established theoretical and observational results .
+ the expressions for the matter - energy density @xmath121 and dark - energy density @xmath122 are given by @xmath123 and @xmath121 versus red shift @xmath98 @xmath77 , @xmath78 , @xmath79 , @xmath80 , @xmath81 , and @xmath82 .
the dots locate the current values of @xmath122 and @xmath121 . in this case , we fix @xmath83.,title="fig : " ] + and @xmath121 versus red shift @xmath98 @xmath77 , @xmath78 , @xmath79 , @xmath80 , @xmath81 , and @xmath82 .
the dots locate the current values of @xmath122 and @xmath121 . in this case , we fix @xmath83 . ] and @xmath124 @xmath125,\ ] ] respectively .
+ adding eqs .
( [ eq24 ] ) and ( [ eq25 ] ) , we obtain total energy ( @xmath126 ) as @xmath127.\ ] ] the variation of the ratio of de and dm energy densities @xmath128 in our model is depicted in fig . @xmath129 . from this figure
we observe that for high red shifts the dm is converting to de whereas by decreasing red shift this conversion reverses due to the interaction between these two dark components . on other word , for small red shifts de being converted to dm which leads to the solution of the coincidence problem .
it is interesting to note that although the de to dm energy transfer happens at smaller red shifts depending on the interaction strength , ultimately i. e at @xmath67 , the @xmath128 ratio tends to the same valve independent to the interaction strength @xmath1 .
the permitted values of @xmath121 and @xmath122 in our model are also shown in fig.@xmath130 . here
, the dots locate the current values of @xmath122 and @xmath121 . from this figure
we see that at late time the total energy density tends to one i.e @xmath131 .
this result is compatible with the observational results .
from both figures @xmath132 we observe that the interaction parameter @xmath1 brings impact on the evolution of the densities depending to its value .
since the investigation of dark energy , there have been many different de models proposed to describe the current cosmic acceleration . to discriminating between these various contenders , sahni
et al ( 2003 ) introduced a new cosmological diagnostic pair @xmath2 called the statefinder in a model - independent manner .
the parameters @xmath133 and @xmath134 are dimensionless and only depend on the scale factor @xmath135 , therefore @xmath2 is a geometrical diagnostic .
the above statefinder diagnostic pair has the following form @xmath136 here we extended the formalism of sahni and coworkers to permit curved universe models .
in fact , trajectories in the @xmath137 plane corresponding to different cosmological models exhibit qualitatively different behaviors . for example , for quintessence and phantom models
the trajectories lie in the region @xmath138 whereas for chaplygin gas models trajectories lie in region @xmath139 .
the quintessence , phantom and chaplygin gas models tend to approach the @xmath140cdm fixed point @xmath141 . + evolution diagram .
the dots locate the current values of the pair @xmath142.,title="fig : " ] + evolution diagram .
the dots locate the current values of the pair @xmath142 . ] in general , the statefinder parameters are given by @xmath143 @xmath144 since from ( [ eq23 ] ) , we have the analytical expression of @xmath95 we can easily obtain @xmath145 .
thus , we can calculate the statefinder parameters in this scenario .
+ the evolution diagrams for the statefinder pair @xmath146 and @xmath142 pair are shown in figures @xmath147 respectively .
the filled circles locate the current values of the pairs @xmath148 and @xmath149 . in fig .
@xmath150 we observe that @xmath133 decreases monotonically to zero , whereas @xmath134 first decreases to a minimum depending to the value of @xmath1 and then rises to unity . in other word , our models tend to approach the @xmath151 which is corresponding to the fixed point @xmath152 at @xmath153 .
@xmath154 shows that all models diverge at the same point in the past @xmath155 and converge to the same point in the future @xmath156 which corresponds to the steady - state models ( ss)- the de sitter expansion ( @xmath157 as @xmath158 and @xmath159 ) . from both figures
@xmath147 , we observe that the interaction parameter @xmath1 makes the model evolve along different trajectories on the @xmath137 and @xmath160 planes .
in this section , we use a direct fitting procedure involving the hubble rate @xmath23 to fix the cosmological bounds on @xmath1 and @xmath4 in our model . to do this
, we use the so called differential age " method proposed by jimenez et al ( 2003 ) . in this method
we can directly study the observational @xmath23 data without using the luminosity distance such as in the case of snia .
+ first of all , we note that the scaled hubble parameter is given by @xmath161^{\frac{1}{2}},\ ] ] where @xmath162 , @xmath121 , and @xmath122 are given by eqs .
( [ eq22 ] ) , ( [ eq24 ] ) , and ( [ eq25 ] ) .
+ following luongo ( 2011 ) we minimize the following reduced @xmath5 in order to constrain the model parameters . @xmath163^{2}}{\sigma^{2}_{hub}(z_{i})},\ ] ] where @xmath164 are the values of table . 1 presented by abraham et al . ( 2004 ) and @xmath165 . in eq .
( [ eq31 ] ) , @xmath166 and @xmath164 are referred to the theoretical and observational values for hubble parameter respectively and the sum is taken over the cosmological dataset .
+ .the cosmological data at 1@xmath1 error for @xmath23 expressed in @xmath167 . [ cols="^,^,^,^,^,^,^,^,^,^,^,^,^,^,^ " , ] [ table : nonlin ] the robustness of our fit can be viewed by looking at fig .
in this paper we have studied the interaction between viscous dark energy and dark matter in the scope of an anisotropic space - time .
it is shown that in absence of interaction , dark matter continuously converts to dark energy whereas interaction leads to an energy transfer from de to dm . therefore ,
as expected , the interaction scenario could lead to the solution of the coincidence problem in cosmology .
we found that either in non - interacting case with small @xmath0 or interacting case with small @xmath1 , the eos parameter only varies in quintessence region which is not in agreement with the recent observation .
we have shown that to get an acceptable scenario of de with an eos parameter evolving from quintessence to phantom and then to the cosmological constant region ( @xmath168 ) one has to consider a viscous de interacting with dm by choosing appropriate values for coupling constant @xmath1 and the bulk viscous coefficient @xmath0 . in this case , our interacting viscous de model does not encounter to the coincidence and big rip problems . moreover , since in our model the viscosity dies out as the universe is expanding , the phantom phase is an temporary state and hence our model will not suffer from the ultraviolet quantum instabilities associated with the phantom models generated by the scalar fields .
it is worth to mention that as argued by carroll et al .
( 2003 ) , any phantom model with @xmath169 should decay to @xmath168 at late time .
in addition , to differentiate between cosmological scenarios involving dark energy , a statefinder diagnostic has been preformed .
this diagnostic indicate that the interaction between de and dm can be probed by the statefinder parameters . finally , we constraint the model parameters @xmath1 and @xmath4 with the observational data for hubble parameter by using chi - squared statistical method .
it is worth to mention that the same method can be used for other bianchi type space - times for which we can find the friedmann - like equations .
the study of interacting viscous de in the scope of other anisotropic bianchi type metrics will be published by the author soon .
this work has been supported by a research fund from the mahshahr branch of islamic azad university under the project entitled interacting viscous dark energy and cold dark matter in an anisotropic universe " .
author also would like to acknowledge the anonymous referee for fruitful comments .
000 abraham , r.g . , et al .
j. * 127 * , 2455 ( 2004 ) alam , u. , sahni , v. , saini , t.d . , starobinsky , a.a .
: mon . not .
. soc . * 344 * , 1057 ( 2003 ) amendola , l. , camargo campos , g. , rosenfeld , r. : phys
d. * 75 * , 083506 ( 2007 ) amirhashchi , h. : astrophys .
space sci .
* 345 * , 439 ( 2013 ) amirhashchi , h. , pradhan , a. , zainuddin , h. : int . j. theor .
phys . * 50 * , 3529 ( 2011 ) amirhashchi , h. , pradhan , a. , saha , b. : astrophys .
space sci . * 333 * , 295 ( 2011 ) amirhashchi , h. , pradhan , a. , saha , b. : chin , phys . lett .
* 28 * , 039801 ( 2011 ) amirhashchi , h. , pradhan , a. , zainuddin , h. : res .
. astrophys . * 13 * , 129 ( 2013 ) astier , p. , et al .
: astron . astrophys . *
447 * , 31 ( 2006 ) balakin , a.b . ,
pavn , d. , schwarz , d.j . , zimdahl , w. : new . j. phys .
* 5 * , 85 ( 2003 ) banijamali , a. , setare , m.r . , fazlpour , b. : int .
. phys . * 49 * , 759 ( 2009 ) barrow , j.d . : class .
quantum grav .
* 21 * , l79 ( 2004 ) benoit , a. , et al .
: astron . astrophys . *
399 * , l25 ( 2003 ) bento , m.c . ,
bertolami , o. , sen , a.a .
rev d. * 66 * , 043507 ( 2002 ) berger , m.s . ,
shojaei , h. : phys .
d. * 74 * , 043530 ( 2006 ) bernardis , p.de .
, et al . :
* 404 * , 955 ( 2000 ) bertolami , o. , et al .
: mon . not .
. soc . * 353 * , 329 ( 2004 ) bertolami , o. , gil pedro f. , le delliou , m. : phys .
b. * 654 * , 165 ( 2007 ) brevik , i. , gorbunova , o. : gen .
37 * , 2039 ( 2005 ) caldwell , r.r .
b. * 545 * , 23 ( 2002 ) caldwell , r.r . ,
kamionkowski , m. , weinberg , n.n .
lett . * 91 * , 071301 ( 2003 ) carroll , s.m . : living . rel . * 4 * , 1 ( 2001 ) carroll , s.m .
, hoffman , m. , trodden , m. : phys .
d. * 68 * , 023509 ( 2003 ) cataldo , m. , cruz , n. , lepe , s. : phys .
* 619 * , 5 ( 2005 ) cimento , l.p . , et al . :
d. * 74 * , 087302 ( 2003 ) copeland , e.j . , sami , m. , tsujikawa , s. : int . j. mod .
d * 15 * , 1753 ( 2006 ) cruz , m. , et al .
: astrophys .
j. * 655 * , 11 ( 2007 ) dalal , n. , abazajian , k. , jenkins , e. , manohar , a.v .
. lett . * 87 * , 141302 ( 2001 ) eckart , c. : phys . rev .
* 58 * , 919 ( 1940 ) feng , b. , wang , x.l . ,
zhang , x. : phys .
b. * 607 * , 35 ( 2005 ) feng , c.j . , zhou , x. : phys .
b. * 680 * , 355 ( 2009 ) guo , z.k . , ohta , n. , tsujikawa , s. : phys .
d. * 76 * , 023508 ( 2007 ) halverson , n.w .
: astrophys .
j. * 568 * , 38 ( 2002 ) hinshaw , g. , et al .
: astrophys .
* 180 * , 225 ( 2009 ) jaffe , t.r . ,
banday , a.j . ,
eriksen , h.k . , g@xmath170rski ,
k.m . , hansen , f.k .
: astrophys .
j. * 629 * , l1 ( 2005 ) jimenez , r. , verde , l. , treu , t. , stern , d. : astrophys .
j. * 593 * , 622(2003 ) jamil , m. , rashid , m.a .
: eur , phys .
j. c. * 58 * , 111 ( 2008 ) jamil , m. , rashid , m.a .
: eur , phys . j. c. * 60 * , 141 ( 2009 ) komatsu , e. , dunkley , j. , et al . : astrophys .
* 180 * , 330 ( 2009 ) landau , l.d.,lifshitz , e.m . : fluid mechanics , 2nd .
, pergamon press , oxford , sect .
49 , ( 1987 ) le delliou , m. , bertolami , o. , gil pedro , f. : aip conf
. proc . * 957 * , 421 ( 2007 ) luongo , o. : mod .
a. * 26 * , 1459 ( 2011 ) mason , b.s . , et al , astrophys .
j. * 591 * , 540 ( 2003 ) mcinnes , b. : j. high energy phys .
* 0208 * , 029 ( 2002 ) miller , a.d . , et al .
: astrophys .
* 524 * , l1 ( 1999 ) nesseris . ,
perivolaropoulos , l. : phys.rev .
d. * 70 * , 123529 ( 2004 ) nojiri , s. , odintsov , s.d .
b. * 562 * , 147 ( 2003 ) oliver f , piattella , jlio c , fabris .
, zimdahl , w. : jcap . * 029 * , 1 ( 2011 ) padmanabhan , t. , chitre , s. : phys .
a. * 120 * , 433 ( 1987 ) padmanabhan , t. : phys . rep .
* 380 * , 235 ( 2003 ) page , l. , et al .
: astrophys .
* 148 * , 233 ( 2003 ) pavon , d. , wang , b. : gen .
relativ . gravit . *
41 * , 1 ( 2009 ) peebles , p.j.e . , ratra , b. : rev .
phys . * 75 * , 559 ( 2003 ) perivolaropoulos , l. aip conf . proc . *
848 * , 698 ( 2006 ) perlmutter , s. , et al .
: astrophys .
j. * 517 * , 565 ( 1999 ) pradhan , a. , amirhashchi , h. , saha , b. : int . j. theor .
* 50 * , 2923 ( 2011a ) pradhan , a. , amirhashchi , h. , saha , b. : astrophys .
space sci . * 333 * , 343 ( 2011b ) ratra , b. , peebles , p.j.e .
: phys . rev .
d. * 37 * , 321 ( 1988 ) riess a.g .
, et al . :
j. * 607 * , 665 ( 2004 ) riess , a.g . , et al .
j. * 116 * , 1009 ( 1998 ) rodrigues , d.c .
d. * 77 * , 023534 ( 2008 ) saha , b. : mod .
a. * 20 * , 2127 ( 2005 ) saha , b. , amirhashchi , h. , pradhan , a. : astrophys .
space sci . * 342 * , 257 ( 2012 ) sahni , v. , saini , t.d .
, starobinsky , a.a . , alam , u. : jetp lett . * 77 * , 201 ( 2003 ) setare , m.r . : eur .
j. c. * 50 * , 991 ( 2007a ) setare , m.r .
j. c. * 52 * , 689 ( 2007b ) setare , m.r .
b. * 654 * , 1 ( 2007c ) setare , m.r . : chin .
* 26 * , 029501 ( 2009a ) setare , m.r . :
. phys . * 49 * , 759 ( 2009b ) setare , m.r . , banijamali , a. , fazlpour , b. : int . j. mod
d. * 20 * , 269 ( 2011 ) setare , m.r .
, sadeghi , j. , amani , r.r .
b. * 673 * , 241 ( 2009 ) setare , m.r .
, sheykhi , a. : int . j. mod .
d. * 19 * , 1205 ( 2010a ) setare , m.r . , sheykhi , a. : int . j. mod
d. * 19 * , 171 ( 2010b ) sheykhi , a. , setare , m.r .
a. * 26 * , 1897 ( 2011 ) sheykhi , a. , setare , m.r .
. phys . * 49 * , 2777 ( 2010 ) singh , c.p .
: pramana .
j. phys . * 71 * , 33 ( 2008 ) spergel , d.n .
: astrophys .
j. * 148 * , 175 ( 2003 ) srivastava , s.k .
b. * 619 * , 1 ( 2005 ) szydlowski , m. , hrycyna , o. : ann .
322 * , 2745 ( 2007 ) tegmark m. , et al .
d. * 69 * , 103501 ( 2004 ) weinberg , s. : rev .
* 61 * , 1 ( 1989 ) wetterich , c. : nucl .
b. * 302 * , 668 ( 1988 ) xu , l. , liu , h. , ping , yo . : int .
. phys . * 45 * , 843 ( 2006 ) zhang , x. : phys .
b. * 611 * , 1 ( 2005a ) zhang , x. : commun .
. phys . * 44 * , 762 ( 2005b ) zimdahl , w. , pavon , d. : gen , relativ , gravit .
* 36 * , 1483 ( 2004 ) zimdahl , w.d . ,
schwarz , j. , balakin , a.b .
, pavn , d. : phy .
d. * 64 * , 063501 ( 2001 ) | a solution to the coincidence and big rip problems on the bases of an anisotropic space - time is proposed .
to do so , we study the interaction between viscous dark energy and dark matter in the scope of the bianchi type - i universe .
we parameterize the viscosity and the interaction between the two fluids by constants @xmath0 and @xmath1 respectively .
a detailed investigation on the cosmological implications of this parametrization has been made .
we have also performed a geometrical diagnostic by using the statefinder pairs @xmath2 and @xmath3 in order to differentiate between different dark energy models .
moreover , we fit the coupling parameter @xmath1 as well as the hubble s parameter @xmath4 of our model by minimizing the @xmath5 through the age differential method , involving a direct measurement of @xmath6 .
+ + + pacs numbers : 98.80.es , 98.80-k , 95.36.+x + key words : bianchi type - i model , dark energy , statefinder |
the ground state maser lines of the oh molecule have long been important sources of information about the regions in which they are formed .
the main lines at 1665 and 1667 mhz and the satellite line at 1612 mhz have been widely observed in star forming regions and the envelopes of late - type stars .
the satellite line at 1720 mhz has been the least commonly observed and has received the least attention until recently .
however , the 1720 mhz maser is now known to be an important indicator of interactions between supernova remnants and molecular clouds .
this connection was first suggested by goss & robinson ( 1968 ) , who noted a new class of 1720 mhz masers in the supernova remnants ( snr ) w28 and w44 .
they detected strong 1720 mhz emission , while the other ground state lines were in absorption .
little additional study of these masers was performed until recently , when the connection between 1720 mhz maser emission and snr s was clearly established by frail , goss & slysh ( 1994 ) . in a followup survey , 1720 mhz maser emission was detected in 17 out of 160 snr s ( green et al .
another source of 1720 mhz maser emission has also been discovered near the galactic center ( yusef - zadeh et al .
1996 ) . all these masers appear to be the result of shock waves propagating perpendicular to the line of sight , hitting adjacent molecular clouds . in their comprehensive study , frail et al . ( 1996 ) note that the maser velocities are usually close to those of the surrounding gas .
this finding is corroborated by claussen et al .
( 1997 ) , who also find in the particular case of ic443 that the masers are confined to one clump of the snr .
earlier molecular line studies by van dishoeck , jansen , & phillips ( 1993 ) showed that , in that particular clump , the shock is propagating perpendicular to the line of sight .
therefore , as could be expected , the maser emission is strongest parallel to the shock front , the direction of largest velocity coherence along the line of sight .
the molecular line studies of van dishoeck et al .
( 1993 ) also find that in ic443 the pre - shock temperature is 15 k and density 4 .
the corresponding postshock values are 80 k and 5 , respectively .
these postshock measurements are supported by the recent h@xmath1co and co observations of frail & mitchell ( 1998 ) , who determined a temperature @xmath2 80 k and densities of in their study of w28 , w44 and 3c391 .
estimates of the post - shock oh column density in w28 range from to @xmath0 ( frail et al .
finally , yusef - zadeh et al.(1996 ) , claussen et al .
( 1997 ) and koralesky et al .
( 1998 ) detected strong polarization and inferred magnetic fields up to a few milligauss in these maser regions .
the basic pumping mechanism for the 1720 mhz masers was identified long ago ( elitzur 1976 ; e76 hereafter ) as collisional excitations at moderate temperatures ( @xmath3 200 k ) , densities ( @xmath4 5 ) and oh column densities ( @xmath0 ) . here
we refine these old calculations with the latest collision rates and use the constraints imposed by the pumping requirements to construct a detailed model for the masers around snr s . in section
ii we discuss the collisional pumping of the 1720 mhz maser and present the results of calculations that show the effects of gas temperature , molecular density and dust radiation on the maser intensity . in section iii
we analyze the constraints the pumping imposes on the type of shock and present a shock model that is in good agreement with the 1720 mhz maser observations .
section iv contains our conclusions .
at the low temperatures in the 1720 maser region , most of the oh molecules populate the four hyperfine levels of the ground state ( @xmath5 ) ( see figure 1 ) .
excitations to the higher rotational levels followed by radiative cascades back to the ground reshuffle the molecules among the ground - state hyperfine levels , leading to inversion of one of the satellite lines when the final cascade is optically thick . when that final decay is from the first excited state , ( @xmath6 ) ,
the 1720 mhz transition is inverted .
however , when the final decay is from the next excited state , ( @xmath7 ) , the 1612 mhz line is inverted instead . as a result
, the 1720 mhz inversion occurs only if the final cascade back to the ground is dominated by optically thick decays from the first excited state .
this condition is met only when excitations are by collisions at @xmath3 200 k , so that ( @xmath6 ) is preferentially populated , and , so that decays from ( @xmath7 ) are optically thin ; when is increased sufficiently , the inversion switches to the 1612 mhz line .
these basic principles have been established long ago ( e76 ) and verified in numerous observations , most notably by van langevelde et al .
( 1995 ) .
the mechanism responsible for the 1720 mhz inversion arises from the basic energy level structure of oh . in particular , it involves simple level counting and is largely independent of the specific form of collision rates ; inversion should occur as long as these rates do not exhibit peculiar selection rules .
due to the lack of calculated collision rates , the e76 study employed hard sphere rates in which all downward collisions have equal strength .
the calculation of actual oh collision cross sections is difficult , and the initial attempts were plagued by serious errors .
these mistakes , reviewed by flower ( 1990 ) , sometimes led to completely opposite conclusions concerning the pumping of the various oh maser lines .
however , the latest collision rates calculated by offer et al .
( 1994 ) for the 24 hyperfine levels shown in figure 1 are more accurate than earlier ones and are also in reasonable agreement with experiment .
the present work utilizes the offer et al .
( 1994 ) collision rates to update the original e76 calculations .
similar to that study , the level populations are calculated in the escape probability formalism and here we employ the expression developed by capriotti ( 1965 ) for a quiescent slab . we have also investigated the effect of dust continuum radiation on the maser strength .
the dust intensity is computed from the approximation @xmath8 b_\nu(t_d)\ ] ] where @xmath9 is the planck function of the dust temperature @xmath10 and @xmath11 is the dust optical depth with a spectral shape following the standard ism dust cross section ( draine & lee 1984 ) .
the calculation of the oh level populations is complicated by the phenomenon of line overlap , which occurs when the frequencies of two or more spectral lines are sufficiently close that photons emitted by one can be absorbed by another .
this affects the escape of the photons and the level populations .
we have incorporated the effects of overlap due to thermal motions with the aid of the method of lockett & elitzur ( 1989 ) .
we compute the populations of the 24 levels shown in figure 1 for a variety of physical parameters , assuming thermal linewidths . from the level populations we find the maser optical depth @xmath12 in the direction of shock propagation .
the largest maser amplification will occur in the direction of largest velocity coherence and will tend to be perpendicular to the shock motion , similar to the case of water masers in star - forming regions ( elitzur , hollenbach & mckee 1989 ) .
our results are presented in figures ( 2)(4 ) , which display the effects of temperature , density and external radiation field on the maser .
the figures show how each property affects the variation of maser optical depth @xmath12 with oh column density .
figure ( 2 ) shows the effect of gas temperature .
strong maser emission constrains the temperature to be between about 50 and 125 k , with maximum inversion occurring for @xmath3 75 k. it is interesting to note that the upper limit on the temperature is affected by line overlap . without overlap , significant inversion would persist beyond 125 k , declining only when @xmath3 exceeds 150 k. figure ( 3 ) shows the effect of molecular density .
the maser can not operate at high densities since quenching by collisions occurs at densities above about 35 .
although the inversion persists as the density decreases , very low densities are not compatible with realistic maser gains , @xmath12 2 , which require 4 .
since the oh abundance relative to h@xmath1 is probably no more than , the corresponding overall column density is 4 . both observations and model calculations discussed below limit the oh shell width to less than 2 cm , thus the overall density in the region must exceed 4 .
our calculations therefore suggest that the range of molecular densities for observable maser emission most likely lie between @xmath13 and 55 .
figure ( 4 ) presents the effect of the dust continuum radiation on the maser strength , showing that dust at a temperature @xmath10 = 50 k has little impact on the maser effect .
however , at @xmath10 = 100 k the dust radiation diminishes significantly the maser inversion even with a minimal amount of dust .
snr s often show strong ir continuum emission which would tend to reduce the 1720 mhz maser inversion .
however , observations indicate relatively low dust emission and temperatures in the maser sources ic443 and 3c391 .
burton et al . ( 1990 ) find that the ir emission from ic443 is dominated by line emission rather than by dust continuum radiation .
when line emission is taken into account , a dust temperature of 45 k is found ( van dishoeck et al .
reach & rho ( 1996 ) have recently measured the ir continuum emission from 3c391 using iso and find a dust temperature of 50 k. figure 4 shows that in these sources the dust radiation will not seriously reduce the maser inversion .
furthermore , the galactic center masers observed by yusef - zadeh et al .
( 1996 ) are located outside the region of intense ir emission ( mezger , duschl , & zylka 1996 ) .
the results presented here are similar to those derived in e76 with hard sphere collisions , corroborating the conclusion of that study that the 1720 mhz inversion reflects the basic oh energy level structure and is largely independent of the exact form of collision rates . because of the severe limitation on oh column density , the maser optical depth is never large .
indeed , even under optimal pumping conditions the 1720 mhz maser barely reaches saturation for radiation propagating along the shock direction ; when @xmath12 is increased to the point that amplified maser intensity is at saturation level , is already so large that the inversion switches to the 1612 mhz maser .
however , significant brightness temperatures can be produced for radiation propagating parallel to the shock front with amplification optical depth @xmath14 , where @xmath15 is an aspect ratio . for
, modest aspect ratios of only 34 suffice to produce brightness temperatures in the range
k. detailed oh pumping calculations based on the offer et al .
( 1994 ) rates were also performed by pavlakis & kylafis ( 1996 ; pk96 hereafter ) , and our results are in general agreement with theirs .
the offer et al .
calculations were the first to include collisions with both ortho- and para - h@xmath1 , and pk96 found that collisions with para - h@xmath1 can eliminate the 1720 mhz inversion .
they suggested , therefore , that the 1720 mhz maser might be used as a tool to determine the ortho - to - para ratio of molecular hydrogen .
our calculations support the pk96 finding that para - h@xmath1 collisions do not lead to 1720 mhz inversion .
however , we conclude that this inversion may not provide a useful indicator of the h@xmath1 ortho - to - para ratio .
first , an examination of the cross sections shows that the dramatic difference in maser efficiency between the ortho- and para - h@xmath16 collisions reflects a minor peculiarity of the cross sections .
para collisions deplete the maser upper level more than the lower level while ortho collisions deplete both at roughly the same rate .
this disparity arises mostly because of a single collision rate : the rate from @xmath17 to the maser s lower level is more than 100 times smaller than that to the upper level for the para collisions . merely setting these two rates equal to each other ,
as they are for the ortho collisions , produces a 1720 mhz inversion also for para - h@xmath16 .
thus the lack of a 1720 mhz maser depends sensitively on just a couple of cross sections .
although the latest rates are the most exact and complete that have yet been calculated , the history of oh collision rate calculations suggests that caution should be exercised in reaching conclusions based on selective collision rates .
a second deficiency of the 1720 mhz maser as a diagnostic of the ortho to para ratio is that it takes just a small amount of ortho to produce the maser since the relevant ortho collision rates are much larger than the para rates .
the equilibrium ratio of ground state ortho to para h@xmath1 is 9exp(@xmath18 ) .
however , the ratio is predicted to be higher than the equilibrium value at low temperatures ( le bourlot 1991 ) , indeed , observations of protostellar outflows yield a ratio of @xmath19 ( smith , davis & lioure 1997 ) .
it is believed that h@xmath1 forms on grains at an ortho to para ratio of 3:1 , and that radiative decays preserve this ratio as the gas cools down .
the results presented in figures ( 2)(4 ) used an ortho to para ratio of 3:1 .
strong 1720 mhz maser emission would persist as long as this ratio is larger than one .
although this ratio will probably influence the strength of the 1720 mhz maser , it would be difficult to estimate its value from the maser intensity alone .
our pumping calculations predict that the 1720 mhz maser should arise under the following conditions : @xmath3 50125 k , @xmath4 5 , .
these parameters are in excellent agreement with the observations ( 1 ) , which clearly establish that the 1720 mhz masers are produced at the shock interface of a snr with a molecular cloud .
it is also significant that these conditions do not result in production of the other ground state masers .
as discussed earlier ( 2 ) , the 1612 mhz maser requires a larger oh column for its creation while the main line masers require high dust temperatures ( elitzur 1992 ) , which would destroy the 1720 mhz maser .
it is also important that a recent search did not detect the 22 ghz water maser in association with the 1720 mhz maser ( claussen et al .
this non - detection is significant , since observable 22 ghz water maser emission is not expected for the densities and temperatures present in these regions .
we now model the environment that can generate these conditions , and show that it requires a c - type shock .
thus the production of the 1720 mhz maser can be used as a diagnostic of the shock structure .
our conclusion that the shock must be c - type is based solely on the physical conditions necessary for 1720 mhz maser action . because of the rapid cooling of the shocked material , the time spent in the necessary temperature range is so short that the accumulated column density is much too small to produce observable maser emission .
we will show that ambipolar diffusion heating in c - shocks can overcome this rapid cooling , but that no equivalent heat source exists for j - shocks .
consider the material leaving a shock front , where it was heated to a high temperature .
as this hot material streams away , it both cools and slows down so that its temperature @xmath3 , velocity @xmath20 , and density @xmath4 are unique functions of distance @xmath21 from the front . assuming an elevated oh abundance of
gives a lower limit of for the h@xmath1 column density of the maser , and the shock must be able to produce such a column for the shocked material as it cools down from @xmath22 125 k to @xmath23 50 k. from mass conservation , the column contained between these two temperatures is ( elitzur 1980 ) @xmath24 where @xmath25 is the pre - shock density , @xmath26 is the shock velocity and @xmath27 is the cooling timescale (= @xmath28 , where @xmath29 is the net energy loss rate per particle ) .
this relation is a direct consequence of mass conservation and applies to all shocks .
its evaluation requires an estimate of @xmath30 , i.e. , @xmath29 .
assume first that there is no heating in the shocked region , as is the case in standard j - shocks , then @xmath29 is determined by the cooling processes only .
detailed calculations show that the dominant coolant in the relevant temperature region is oi , through its @xmath31 fine structure transitions ( e.g. hollenbach & mckee 1989 ) , and simple estimates show that @xmath32 6@xmath3 ergs@xmath33 per h@xmath1 in this case .
this gives a temperature - independent cooling timescale 68 s , an upper limit on @xmath30 since other coolants may also contribute .
thus the previous equation becomes @xmath34 where @xmath35 and @xmath36 . to support the 1720 mhz maser
this column must be , therefore the ambient density and shock velocity are constrained by @xmath37 now , in a j - shock , to a good degree of approximation the ram pressure ( @xmath38 ) is equal to the thermal pressure of the shocked material ( elitzur 1980 ) . since the pumping requirements place upper limits on both the density ( @xmath4 55 ) and temperature ( @xmath3 125 k ) of the maser region , it follows that its thermal pressure is bounded by @xmath39 6.257k erg .
therefore , pressure balance across a j - shock yields @xmath40 another constraint on @xmath25 and @xmath26 .
the last two constraints can be obeyed simultaneously only if @xmath26 33 cms@xmath33 and @xmath25 38 , and these two bounds are both unphysical .
the upper bound on the shock velocity is two orders of magnitude less than the speed of sound , and the _ lower _ limit on the _ pre - shock _ density exceeds the _ upper _ limit on the _ post - shock _ maser density by more than three orders of magnitude .
theses difficulties persist for j - shocks even when the post - shock pressure is dominated not by the thermal motions but rather by a frozen - in magnetic field . assuming flux freezing and an ambient field @xmath41 g , pressure - balance across the shock yields @xmath42 , where @xmath43 is the post - shock density ( e.g. hollenbach & mckee 1989 ) . together with equation ( [ constraint1 ] ) this implies @xmath4 7 , a lower limit exceeding the upper bound on maser density by more than an order of magnitude .
j - shocks can not produce the relatively high column densities necessary for observable maser emission at the relatively low thermal pressures dictated by the pumping requirements of the 1720 mhz maser .
the failure can be traced directly to the low column densities that these shocks produce in the relevant temperature region . from equation
[ column ] , typical ambient densities and shock velocities produce column densities that are more than two orders of magnitude smaller than required .
the problem is that @xmath30 is so short , since the cooling is so fast , and can be circumvented only if some heating existed to counter the cooling and maintain the temperature in the 50125 k range for a longer period , extending the column .
one possible heat source in j - shocks is the heat of h@xmath1 formation on grains , which can be tapped if the shock is sufficiently fast to first dissociate all molecules ( hollenbach , mckee , & chernoff 1987 ) .
indeed , this is the heating mechanism invoked to create the right conditions for water masers in dissociative j - shocks with pre - shock densities 36 ( elitzur et al . 1989 ) .
however , since this mechanism involves collisional de - excitation of h@xmath1 vibrational states , its efficiency deteriorates so rapidly at lower densities and temperatures that it becomes inoperable at the conditions required here .
another possibility is radiative heating .
one common method is to first radiatively heat the dust , transferring the heat to the gas by collisions .
this mechanism is not feasible here because the radiation from the heated dust would destroy the maser inversion .
another radiative heat source could be ionization by the snr x - rays , since a significant fraction of the ionization energy goes into heating .
maloney , hollenbach & tielens ( 1996 ) estimate that 3040% of the ionization energy goes into heating the gas if it is primarily molecular . from the expressions they provide ,
the x - ray heating rate for the conditions considered here is only @xmath44 ergs@xmath33 per h@xmath1 , about 4 orders of magnitude less than the cooling rate and thus a negligible heating source .
heating of the shocked material can be provided by ambipolar diffusion .
this mechanism occurs when flux - freezing breaks down and the shock then switches to c - type . in the parameter regime
relevant here , especially the magnetic fields inferred from the polarization measurements , this is the expected shock type .
indeed , in their analysis of line emission from ic443 , burton et al .
( 1990 ) concluded that c - shocks are present in that source .
furthermore , reach and rho ( 1998a ) have recently found that the brightness of the co lines they observed in the 3c391 maser region is consistent with c - type molecular shocks with @xmath45 @xmath46 and @xmath47 .
detailed c - shock calculations were performed by draine , roberge and dalgarno ( 1983 , drd hereafter ) .
they present results for a number of models that bracket the general requirements for 1720 mhz maser operation .
in particular , their intermediate density model has @xmath25 = 4 and @xmath26 = 25 .
this shock develops a region with the appropriate temperature ( 50125 k ) and density ( 5 ) that extends over a distance of cm , i.e. , a column density of .
these are just the parameters required for the 1720 mhz maser .
unfortunately , this potential maser region has virtually no oh ; as is well known , molecular shocks channel into h@xmath1o , rather than oh , essentially all the oxygen not in co , and this is the case also with the drd model .
however , subsequent photodissociation , which was not included in the drd calculations , can convert the water into oh .
such dissociation could be produced by the external uv radiation field as in the case of shocks around hii / oh regions ( elitzur and de jong 1978 ) .
the relevant photodissociation rates were listed most recently by roberge et al .
( 1991 ) @xmath48 here @xmath49 is the local enhancement of the standard interstellar uv field ( draine 1978 ) and @xmath50 is the source s optical depth at visual wavelengths . with these rates we can solve analytically for the time dependent oh and h@xmath1o abundances of any parcel of gas after its exposure to uv radiation . note that time @xmath51 enters into the equations only in the combination @xmath52 .
the solution for a mixture that is initially purely h@xmath1o is displayed in figure ( 5 ) , which shows the oh abundance as a fraction of the gaseous oxygen not in co. another dissociation mechanism , recently advocated by wardle et al .
( 1998 ) , involves the secondary uv photons produced after x - ray absorption . with expressions from maloney et al .
( 1996 ) , this water dissociation rate is @xmath53 where @xmath54 erg s@xmath33 is the snr x - ray luminosity , @xmath55 pc is the snr radius and @xmath56 is the hydrogen column attenuating the x - ray flux .
the three snrs ic443 , w44 and 3c391 have @xmath57 .12 , @xmath55 10 and @xmath58 0.11 ( charles & seward 1995 ; rho et al . 1994 ; rho & petre 1996 ) .
these parameters yield a water dissociation rate of sec@xmath33 , about 1000 times smaller than that due to the standard ism field . therefore this mechanism can be neglected , unless the x - ray flux has been underestimated by more than a factor of 50 .
the width of the maser region in the drd model is cm . with a shock speed of 25 ,
the corresponding crossing time is 125 years , same as the time of the peak evident in figure 5 .
thus this figure suggests that maximum oh abundance would occur for @xmath59 , the strength of the unattenuated standard interstellar field .
however , radiation this intense would significantly increase the electron abundance and could drastically affect the shock structure to the point that it is no longer a c - shock .
drd note that a modest change in the pre - shock electron density will not significantly affect the structure of their 4 c - shock because most of the momentum transfer is via charged dust grains .
however , the standard , unshielded interstellar field would result in an increase of the electron abundance from to about , and such a large increase would have a significant effect on the shock structure .
thus there are two competing constraints on the strength of the uv field : it must be strong enough to create sufficient oh , yet not so strong as to increase the electron abundance too much .
a field 100 times weaker than the standard interstellar field appears to be a reasonable choice .
it will increase the electron abundance by no more than an order of magnitude while producing the required amount of oh . with @xmath60 , the time axis in figure 5 is stretched so that roughly 10% of the water is converted to oh in the maser region and the oh abundance is about 2 .
these results are in good agreement with the recent observations of reach and rho ( 1998b ) who detected both oh and h@xmath1o at an abundance ratio of @xmath2 1:15 in a shocked region in 3c391 .
although more accurate estimates of the oh abundance will require shock calculations that directly include photodissociation effects , this analysis shows that water photodissociation can lead to the necessary oh column density without serious effects on the shock structure .
finally , the ambipolar diffusion process is driven by magnetic field gradients . in a c - shock
these gradients are generated by the shock compression .
additionally , gradients can be produced by field curvature , an attractive possibility given the apparent maser location at the edge of an expanding snr .
a gradient characterized by length scale cm would produce ambipolar diffusion whose heating rate is similar to that of the drd c - shock model .
such a gradient , which could be reasonably expected at the interface of clumps overtaken by the snr shell , would augment the c - shock heating effect even if the ionization were higher than what drd assumed in their calculation .
for many years the 1720 mhz transition was the least observed of the oh ground state maser lines .
our work shows that the inherent weakness and rarity of this maser reflect a combination of independent factors .
first , because of the severe limitation on oh column density , the 1720 mmz maser optical depth is never large . even when produced under optimal conditions , appreciable maser amplification would require radiation propagation parallel to the shock front and sizeable aspect ratios .
second , the production of this maser places tight constraints on the physical environment .
furthermore , the particular combination of physical parameters necessary for this maser is quite uncommon and only c - shocks seem capable of producing it .
third , even c - shocks are incapable of generating the optimal conditions for this maser .
the oh columns produced by these shocks correspond to @xmath12 smaller than its potential peak value by more than a factor of 2 , a significant factor in light of the steep dependence of maser emission on optical depth .
thus our work provides a natural explanation for the low detection rate of 10% found by green et al .
( 1997 ) . while the tight constraints on the physical conditions needed for its production make for a weak 1720 mhz maser , they also make it a powerful probe of the region where it is formed .
our most significant result is that the detection of the 1720 mhz maser rules out the possibility of a j - shock ; only c - shocks can produce the conditions required for this maser .
this conclusion , too , is supported by the detection of high maser polarization and the magnetic fields it implies . because of the wide margin by which j - shocks fail to produce it , the 1720 mhz maser could be one of the best discriminators of the shock type , a reliable indicator of the presence of a c - shock .
we thank e. van dishoeck for sending us the tabulations of the offer et al .
collision rates , and d. frail , d. hollenbach and w. reach for most useful comments on the original manuscript . the partial support of nasa grant nag5 - 7031 is gratefully acknowledged .
after completion of this work we have learned of a study by wardle , yusef - zadeh & geballe ( 1998 ) that reached similar conclusions regarding the role of c - shocks for these masers .
in the absence of external radiation , the level populations are determined by the following parameters : overall density @xmath4 , @xmath61 and the temperature .
when the pumping is collisional and all the relevant lines are optically thick , the first two parameters enter only in the combination @xmath62 a scaling parameter that proved useful in the analysis of water maser pumping ( elitzur et al .
since the 1720 mhz maser , too , is collisionally pumped , we investigated the scaling properties of its inversion efficiency @xmath63 .
we find that to a good degree of approximation , @xmath64 is indeed a function only of @xmath3 and @xmath65 , i.e. , @xmath4-independent , when @xmath65 @xmath66/ and @xmath67 35 . at smaller values of @xmath65
we find that @xmath64 remains independent of @xmath4 but @xmath65 is replaced by @xmath61 as the relevant scaling variable .
both scaling properties can be verified analytically with the aid of a simple 8-level model of the first two rotation states , employing hard - sphere collision rates . in this model
the inversion efficiency obeys @xmath68 where level numbering is in order of increasing energy ( see figure 1 ) , @xmath69 is the radiative decay rate and @xmath70 is the ( uniform ) downward collision rate .
note that @xmath70 depends only on @xmath3 and @xmath4 while @xmath71 depends only on @xmath61 . at low densities and column densities
@xmath72 , therefore @xmath64 depends only on @xmath71 , i.e. , @xmath61 .
the dependence on @xmath61 persists as long as some of the radiative transitions remain optically thin . as the column density increases ,
the radiative decay rates become smaller , due to photon trapping , and downward collisions begin to dominate .
eventually all the relevant optical depths become large and the maser efficiency becomes a function of @xmath73 , i.e. , it scales with @xmath65 .
thanks to scaling , the numerical results presented here can be extended to other densities . however , because of its small phase space , the usefulness of scaling for the 1720 mhz maser is more limited than for water masers . | recent observations show that the oh 1720 mhz maser is a powerful probe of the shocked region where a supernova remnant strikes a molecular cloud .
we perform a thorough study of the pumping of this maser and find tight constraints on the physical conditions needed for its production .
the presence of the maser implies moderate temperatures ( 50 125 k ) and densities ( 5 ) , and oh column densities of order @xmath0 .
we show that these conditions can exist only if the shocks are of c - type .
j - shocks fail by such a wide margin that the presence of this maser could become the most powerful indicator of c - shocks .
these conditions also mean that the 1720 mhz maser will be inherently weak compared to the other ground state oh masers .
all the model predictions are in good agreement with the observations .
# 1 |
the @xmath0-axis transport in the highly anisotropic cuprate superconductors tl@xmath4ba@xmath4ca@xmath4cu@xmath5o@xmath6 ( tbcco ) and bi@xmath4sr@xmath4cacu@xmath4o@xmath7 ( bscco ) can well be described by a stack of josephson junctions between the superconducting cuo@xmath4-multi - layers .
this intrinsic josephson effect can be seen in the multi - branch structure of the @xmath2-@xmath3 characteristic , each branch corresponding to a well - defined number of junctions in the resistive state.@xcite the intrinsic josephson effect is observed also in the behaviour of the material in external magnetic fields and under microwave irradiation.@xcite recently subgap structures in the @xmath2-@xmath3-characteristic have been discovered as intrinsic properties of the material,@xcite which could be explained by the coupling between the intrinsic josephson oscillations and phonons.@xcite this interaction is mediated by the oscillating electric field in the josephson junction , which excites vibrations of charged ions in the material . in our previous investigations @xcite
we used a simple model of a system of damped harmonic oscillators in order to describe the dynamics of oscillating ions in the barrier .
we were able to derive an analytic expression for the dc - current density @xmath8 as function of the dc - voltage @xmath3 for one resistive junction @xmath9 where the voltage @xmath3 is related to the josephson oscillation frequency @xmath10 by @xmath11 and @xmath12 is the dielectric function of the oscillating ions . from this result
it can be seen that the @xmath2-@xmath3-curve has a maximum at the frequency ( voltage ) where the real part @xmath13 of the phonon - dielectric function vanishes , which corresponds to a longitudinal eigenfrequency of the phonon system . with appropriate values for the josephson plasma freqency @xmath14 , the quasiparticle conductivity @xmath15 , the critical current denstiy @xmath16 , and the frequencies , dampings and oscillator strengths of phonons in the dielectric function we were able to fit the experimental results for the sub - gap structures in the @xmath2-@xmath3-curves perfectly .
in addition to this , the identification of the maxima of the structure with phonon frequencies provides a natural explanation , why the position of the resonances is completely independent of temperature , magnetic field and the geometry of the probe .
the voltage positions of these resonances can also be related to structures in optical experiments,@xcite in particular to the reflectivity for oblique incidence@xcite . despite the clear evidence of the effect a general theory for the coupling between josephson oscillations and phonons is still missing and
shall be presented in the following . in particular , in refs .
the phonon frequencies and the oscillator strengths in the dielectric function were not specified .
the clarification of this point and the extension to a general lattice dynamical model including acoustical and optical branches will be one of the main topics of the following investigation .
another topic concerns the coupling of josephson oscillations in different resistive junctions .
we will show how phonons lead to a coupling of the phases of josephson oscillations in different resistive junctions .
this will be discussed in detail for two resistive junctions and general results will be given for large stacks of resistive junctions .
a phase - locking in a stack of josephson junctions is important for applications of such systems for high - frequency mixers and detectors .
the excitation of phonons by josephson oscillations in conventional single josephson junctions has been observed already a long time ago.@xcite also in the @xmath2-@xmath3 curves of break - junctions of cuprate - superconductors @xcite structures due to phonons might have been identified .
the physical mechanism described here can also be applied to these cases but our formalism is particularly suited to treat stacks of josephson junctions with phonons in the frequency range between the josephson plasma frequency and the gap - frequency .
it is not our intention to explain the details of the @xmath2-@xmath3-characteristic of tbcco and bscco by a realistic lattice dynamical calculation .
this is impossible at the moment for the complicated anisotropic superconductors with variable doping showing this effect .
furthermore this would require a detailed theory of superconductivity and the josephson effect in two- and three - layer systems including information about the superconducting bands , charge distribution , and charge susceptibility between the layers and inside the cuo@xmath4 planes which is not available yet .
therefore we discuss a simple model system with superconducting mono - layers where the conduction electron charge is distributed homogeneously along the layers , and a lattice dynamical model with only two phonon bands showing already the basic features of a full theory which will be expected also for realistic systems .
we start the discussion with a definition of our model and a short derivation of the basic josephson equations for a stack of josephson junctions .
we consider a system of @xmath17 superconducting layers separated by insulating barrier material of thickness @xmath18 forming a stack of josephson junctions .
we treat the superconducting layers as homogeneous metal sheets with a uniform electron distribution along the layers . in this paper
we treat only the case of a uniform tunneling current with a constant bias - current and neglect magnetic field effects due to the current flow . in this case
all quantities are constant along the layers .
such an approximation is reasonable for a stack of junctions which is short with respect to the josephson penetration length but long enough to neglect finite size effects in the ionic polarisation .
the tunneling current density @xmath19 from layer @xmath20 to @xmath21 creates ( two - dimensional ) charge density fluctuations @xmath22 on the layers related by the continuity equation @xmath23 these charge fluctuations create electric fields @xmath24 ( in the c - direction ) in the barrier between layer @xmath20 and @xmath21 which are constant inside each barrier and are related to the charge fluctuations by the maxwell equation : @xmath25 or @xmath26 assuming that the current density @xmath19 for the first and last barrier is fixed by the bias current density @xmath27 , then with help of eqs .
( [ cont],[maxwell1 ] ) the tunneling current density @xmath19 in all the other junctions is related to the bias current density @xmath27 by @xmath28 the last term is the displacement current density related to the charge fluctuations on the layers . in the following
we denote this term by @xmath29 in order to relate the present results to the usual notation of the resistively shunted josephson junction ( rsj ) model , however , one should keep in mind the microscopic origin of this term . in the following we approximate the tunneling current by a superposition of a josephson supercurrent density and a quasiparticle current density .
then we have for each junction : @xmath30 the josephson current density @xmath31 depends on the gauge invariant phase difference @xmath32 between layers @xmath20 and @xmath21 at positions @xmath33 and @xmath34 .
it is related to the average total electric field in the barrier @xmath35 by the second josephson equation @xmath36 here small corrections to eq .
( [ 2jos ] ) for layered superconductors which are discussed in refs . are neglected . for the quasi - particle current density we will use in the following an ohmic form @xmath37 with a constant conductivity @xmath15 .
the generalisation to more realistic forms @xcite is straightforward .
the crucial point where the phonons come into play is the relation between the field @xmath38 , which is created by the charge fluctuations on the superconducting layers alone , and the average electric field eq . ( [ totale ] ) @xmath39 which contains also the averaged field @xmath40 created by the ionic displacements in the barrier .
this will be discussed in detail below . before we do this
let us summarize the most important parameters which characterize the josephson system : the first one is the ( bare ) josephson plasma frequency @xmath14 defined by @xmath41 the second one is the so - called characteristic frequency defined by @xmath42 here @xmath43 is the voltage where the quasiparticle current density equals the value @xmath16 .
it is of the order of the superconducting energy gap and is a measure of the dissipative properties of the junction . in our simple model with a constant conductivity
we have @xmath44 .
the ratio @xmath45 is the mccumber parameter , which is @xmath46 for the strongly anisotropic cuprate - superconductors .
moreover , for these materials there exist phonons with frequencies in the range @xmath47 . typically the time - dependence of the phase difference @xmath48 can be written in the so - called resistive state as @xmath49 where @xmath50 is the time - average of the phase velocity which is non - zero for a junction in the resistive state .
it determines the dc - voltage @xmath51 across the junction . in the asymptotically stable state and for large values of the mccumber parameter
@xmath52 the oscillating part @xmath53 is small and oscillates with the frequency @xmath10 .
now let us turn to the discussion of lattice vibrations .
quite generally the lattice displacement of an ion of type @xmath54 with mass @xmath55 , charge @xmath56 in unit cell @xmath57 is determined by the following equation of motion @xmath58 here @xmath59 is the local driving field at the equilibrium position @xmath60 of the ion generated by the charge fluctuations @xmath61 on the superconducting layers . note that the vibrating ions may be both in the barrier material and on the superconducting layers .
the superconducting electrons are assumed to move together with the ions of the layers . while the rsj - equations for the phases are highly non - linear , the relations between lattice displacements and electric fields are linear , consequently we may analyse the response for each frequency @xmath10 separately . with a harmonic ansatz of the form
@xmath62 we obtain for the amplitude @xmath63 here @xmath64 and @xmath65 are the eigenvalues and eigenvectors of the dynamical matrix @xmath66 defined by @xmath67 the force - constant matrix contains the quantum - mechanical short - range interactions but also the short - range and long - range coulomb interactions ( the latter being of the form @xmath68 ) between the ions , but not the fields set - up by the conduction electrons on the layers .
the eigenfrequencies @xmath69 are therefore by construction the _ bare _ phonon frequencies in the absence of the conduction electron charge fluctuations @xmath61 .
they include possible renormalisation by intra - atomic electronic polarisation . for the further discussion it is convenient to label the lattice dynamical unit cells by @xmath70 with @xmath71 denoting the superconducting layer
in which the lattice cell is contained ( see fig .
[ fig1 ] ) . then the z - component of the position vector @xmath72 for @xmath73 does not depend on @xmath74 and we may write @xmath75 .
furthermore the origin of the unit - cell may be chosen on the superconducting layer , so that @xmath76 . in our model for the superconductivity in the layers we have to assume that the mobile electronic charge is spread out uniformly along the layers , because a microscopic theory connecting the superconducting bands and the atomic sites in the cuo@xmath4-layers is still missing .
therefore the field @xmath77 does not depend on the position @xmath78 along the layer and has only a z - component . in eq .
( [ displacement ] ) only modes with @xmath79 appear and we may write for the displacement amplitude in z - direction of an ion of type @xmath54 in barrier @xmath20 : @xmath80 now let us specify the driving field @xmath81 for the ions in more detail : summing up the electric fields created by the charge fluctuations on the different layers we find : @xmath82 note that for ions _ inside _ the barrier the driving field does not depend on the position @xmath83 of the ion and is equal to the constant field @xmath84 introduced in eq .
( [ maxwell2 ] ) : @xmath85 in order to find a relation between the lattice displacements and the driving field we introduce the fourier transformation @xmath86 then @xmath87 with the @xmath88-dependent effective charge @xmath89 and consequently @xmath90 thereby the vanishing of @xmath91 for @xmath92 and for ions on the layers reflects the fact that for alternating electric fields in the barriers no net force is acting on the superconducting layers . next we want to relate the driving field to the average total electric field @xmath93 in the barrier because this field is connected with the phase difference @xmath94 by the second josephson - equation ( [ 2jos ] ) .
the microscopic total electric field @xmath95 is the sum of the fields set - up by both the charge - fluctuations and the displaced ionic charges : @xmath96 .
as @xmath97 is assumed to be constant inside the barrier and independent of x we may replace the other two fields also by their averages across the barrier and over one unit cell along the layer .
for the averaged fields we have then the relation : @xmath98 the field @xmath40 in the barrier can be calculated from the difference of the scalar potentials on layers @xmath20 and @xmath21 produced by the ionic displacements averaged over the area of one unit - cell as @xmath99 here @xmath100 is the volume of one unit cell .
the factor @xmath101 results from the fact that displacements of ions ( with @xmath102 ) on a layer do not contribute to the potential on the same layer .
note that this field for @xmath79 does not depend on the ionic displacements in other barriers and is closely related to the ionic polarisation in the same barrier . defining a generalised polarisation by @xmath103 we may write for the displacement in the barrier @xmath104 which has the usual form as in the macroscopic maxwell theory .
going over to a fourier transformation the relation @xmath105 between the polarisation and the driving field is obtained with @xmath106 and the oscillator strength @xmath107 the special combination of phase factors contained in @xmath108 eq .
( [ zt ] ) are a consequence of the different contribution of ions on and between the superconducting layers to the electric field in the barrier .
using @xmath109 we can solve for @xmath110 : @xmath111 the relation @xmath112 defines an effective longitudinal dielectric function @xmath113 this function has zeros at the eigenfrequencies @xmath114 of the dynamical matrix . due to the form of the oscillator strengths eq .
( [ osc ] ) only modes with polarization in the c - direction contribute . as the electric field @xmath93 is constant along the layers we have @xmath115 and only longitudinal modes in the c - direction couple , therefore the zeros of @xmath116 are exactly at the longitudinal eigenfrequencies of the dynamical matrix . in the case of a single dispersionless phonon - mode with frequency @xmath117 the function @xmath118 can be directly compared with the dielectric function used in ref . .
in fact , in this case @xmath118 can be written as @xmath119 the form of the longitudinal dielectric function @xmath120 eq .
( [ ephon ] ) , which we have introduced here , is different from the transverse dielectric function @xmath121 in both functions different eigenfrequencies and oscillator strengths enter , however in the limit @xmath122 , which is only relevant in optical experiments , the values of both functions are equal .
finally a comparison of our theory with theoretical investigations in ref .
are in order . in principle
there are two different electron - phonon coupling mechanisms , which may couple josephson oscillations and phonons : 1 .
the electromagnetic interaction between the ionic charges and the charges of conduction electrons , 2 .
the dependence of the tunneling matrix element on lattice displacements .
the first mechanism is considered in our work , the second in ref . .
both mechanisms require a different theoretical treatment ( on a diagrammatical basis the two mechanisms would correspond to different diagrams ) .
it has been argued in ref .
that in the layered cuprate superconductors the charges of the ions in the insulating barrier between superconducting layers are unscreened and therefore have a strong interaction with conduction electrons in the cuo@xmath4-layers .
we therefore considered this mechanism for our treatment of the coupling between josephson oscillations and phonons .
though we did not write down a hamiltonian for the interacting system our method nevertheless is a full microscopic theory which treats the electron - phonon interaction on a random - phase - type level by describing the interaction with the help of internal fields .
this approximation is sufficient as long as we do not want to consider the electron - phonon interaction inside the superconducting layers and treat exchange effects between different superconducting layers .
according to the rsj - like model derived in eq .
( [ rsj0 ] ) the current density in junction @xmath20 is @xmath123 where @xmath124 is the electric field in junction @xmath20 set - up by the charge fluctuations of conduction electrons . as pointed out in the previous section this field
can be expressed by the average electric field in the barrier and the generalized polarisation eq .
( [ pol ] ) as @xmath125 .
the polarisation has to be calculated self - consistently from the ionic displacements and depends linearly on the electric field .
let us discuss first the case of one resistive junction at @xmath126 in the middle of a large stack while all other junctions @xmath127 are in the superconducting state .
then as mentioned previously all the oscillations are governed by one frequency @xmath10 , and we can write for the phase for @xmath126 : @xmath128 while for @xmath127 we have @xmath129 in the stationary state @xmath130 oscillates with the same frequency @xmath10 , @xmath131 higher harmonics can be neglected for @xmath46 . in this limit the fluctuations @xmath130 are small and we may use the expansion @xmath132 while for @xmath133 we have @xmath134 the bias current density @xmath27 on the l.h.s . of the rsj - equations eq .
( [ rsj1 ] ) is time - independent and equal for all junctions , while the quantities on the r.h.s have both time - independent and oscillating components .
let us discuss the equations for the non - resistive junctions ( @xmath127 ) first . here
the dc - component is : @xmath135 this fixes the constant part of the phases in the non - resistive junctions and relates it to the bias current . for the oscillating part of eq .
( [ rsj1 ] ) one obtains : @xmath136 or @xmath137 with the reduced josephson plasma frequency @xmath138 now we discuss the resistive junction at @xmath126 . keeping only the lowest harmonic we find @xmath139 the dc - component of the rsj - equation eq .
( [ rsj1 ] ) is therefore given by @xmath140 where @xmath3 is the dc - voltage of the resistive junction and @xmath141 is the quasiparticle current density .
for the oscillating part one finds : @xmath142 the two equations ( [ nne0 ] , [ ne0 ] ) can be combined to one inhomogeneous linear differential equation for all @xmath20 @xmath143 with @xmath144 assuming a time dependence of the form @xmath145 for all oscillating quantities we have @xmath146 with @xmath147 in order to incorporate the non - local dependence of the polarisation on the electric fields in different barriers a spatial fourier representation of the form @xmath148 is introduced .
using the relation @xmath149 in eq .
( [ inh ] ) yields @xmath150 with @xmath151 for the phase oscillation in the resistive junction follows @xmath152 with @xmath153 solving for @xmath154 @xmath155 we obtain @xmath156 from this finally the following expression for the dc - current density as function of the dc voltage is obtained : @xmath157 here @xmath158 is a modified dielectric function @xmath159 where @xmath160 and @xmath161^{-1}. \label{j}\ ] ] this expression describes the dc - current density as function of the dc - voltage @xmath162 .
it has a maximum for frequencies @xmath10 where the real part @xmath163 vanishes , i.e. for longitudinal phonon frequencies .
this can easily be seen if we consider the special case of one phonon mode without dispersion .
then @xmath164 , which is of the form eq . (
[ eps ] ) .
the corresponding @xmath2-@xmath3-curve is shown in fig .
[ fig2 ] , which is calculated with the dielectric function @xmath165 here an additional damping @xmath166 of the phonon has been introduced in eq .
( [ eps ] ) in order to simulate the energy transfer into other junctions due the coupling of the ions .
a peak in the @xmath2-@xmath3-curve appears at @xmath167 .
the width of the peak is determined by this damping together with the quasiparticle conductivity .
the deviation from the quasiparticle current density vanishes at the pole of @xmath118 at @xmath168 .
the rise at low voltages is due to the plasma resonance . for @xmath169
the minimum of the @xmath2-@xmath3-curve is at @xmath170 . in current - biased experiments and corresponding numerical simulations the parts with negative differential conductivity
are skipped hysteretically as indicated in the figure .
note that the denominator in the function @xmath171 eq .
( [ j ] ) is the total @xmath88- and @xmath10-dependent longitudinal dielectric function of the coupled system of phonons and conduction electrons , @xmath172 zeros of the real part of this function describe longitudinal collective modes in the system . on the other hand
the resonances in the @xmath2-@xmath3 curve appear at the _ bare _ longitudinal phonon frequency in the case of a narrow phonon band .
the summation over @xmath88 in eq .
( [ j ] ) leads to an effective damping of the resonances which is proportional to the phonon dispersion .
the physical origin is the loss of energy by phonons from the resistive junction to the neighboring junctions .
the result for the current - voltage characteristic can be generalised to the case of several junctions being in the resistive state , if we assume that all junctions oscillate with the same frequency @xmath10 .
denoting the subset of indices for the resistive junctions by @xmath2 then for @xmath173 we obtain ( for a derivation see the appendix ) : @xmath174^{-1 } e^{-i\theta_{k } } \end{aligned}\ ] ] the dielectric function @xmath175 is defined by @xmath176^{-1 } \\
\nonumber & + & ( \frac{\bar\omega^2_p}{\omega^2 } - \frac{i\sigma}{\epsilon_0 \omega})\delta_{i , k } \end{aligned}\ ] ] the terms in brackets in eqs . ( [ multi1 ] , [ multi2 ] )
are understood as matrix inversions .
the dc - voltage @xmath3 is obtained , if one multiplies @xmath177 by the number of resistive junctions .
note that the r.h.s .
( [ multi1 ] ) depends on the layer index @xmath173 , while the l.h.s .
is equal for each junction . from this equality the ( relative )
phases @xmath178 in the different junctions can in principle be determined , which in turn provides an analytical expression for the @xmath2-@xmath3-curve . in the case of two resistive junctions
exactly two solutions exist with @xmath179 and @xmath180 , respectively . in the general case
several different solutions are found .
the stability of these solutions will be checked by a comparison with a direct numerical integration of the coupled equations of motions in the following section .
it turns out , that those analytical solutions are most stable where the phases @xmath178 of the oscillating josephson junctions fit best to the pattern of lattice vibrations with the given frequency @xmath10 .
the theory in the preceding sections is developed for general lattice dynamical models and is valid within the assumptions we have made for the superconducting properties : we treat only single - layer systems assuming a homogeneous conduction electron charge distribution along the layers .
an extension of the theory to more realistic systems is in principle possible , but requires to introduce the charge susceptibility of conduction electrons in the superconducting state and a generalisation of the josephson theory to multilayer systems . but
for this more details of the electronic properties are required than are currently known about these materials .
in addition to this , it is not possible to compare theoretical results for the @xmath2-@xmath3-curves with experiments in detail , as reliable lattice dynamical calculations for the strongly anisotropic systems bscco and tbcco with two and three layers and variable oxigen content are not yet available . in the following we consequently only want to illustrate the main features of our theory in a simple toy model , which reflects some basic aspects of the real system .
one of the main lattice dynamical property of these systems is certainly the existence of a longitudinal acoustical and ( several ) flat longitudinal optical bands which result from movements of groups of ions in the barrier against ions in the cuo@xmath4-planes in the c - direction.@xcite such modes we simulate by the most simple lattice dynamical model consisting of two kinds of ions with ionic charges @xmath181 , @xmath182 and masses @xmath183 ,
@xmath184 . the first kind ( @xmath185 ) is placed on the superconducting layers , the second kind ( @xmath186 ) in the middle of the barrier .
the motion of ions in the c - direction which is assumed to be uniform along the layers is approximated by a two - atomic chain - model with nearest - neighbor interactions in the c - direction : @xmath187 by choosing the masses very different a narrow optical band can be simulated . from a diagonalisation of the dynamical matrix given by eq .
( [ vib ] ) the well - known eigenfrequencies @xmath188 of the two - atomic chain are obtained . with help of the eigenvectors the oscillator strengths defined in eq .
( [ osc ] ) are calculated , which are needed for the longitudinal dielectric function eq .
( [ ephon ] ) . the driving field on the r.h.s .
( [ vib ] ) is the field set - up by the conduction electron charges on the superconducting layers , which can be expressed by the ( constant ) field @xmath84 in the barrier between layers @xmath20 and @xmath21 with the help of eq .
( [ erho ] ) .
the latter can be expressed by the average total electric field @xmath93 in the barrier : @xmath189 where the polarisation eq .
( [ pol ] ) is given by : @xmath190 these equations for the motion of lattice displacements have to be supplemented by the extended rsj - equations eq .
( [ rsj1 ] ) : @xmath191 this model is used to calculate the @xmath2-@xmath3-characteristics in two ways : 1 .
the @xmath2-@xmath3-curves are calculated analytically using the results eq .
( [ multi1 ] ) obtained by the green s function method .
thus the peaks due to the phonon resonances are obtained .
2 . the coupled set of rsj equations eq . ( [ rsj ] ) and phonon equations eqs .
( [ vib ] , [ pol1 ] ) are integrated numerically by a runge - kutta method for a finite stack of josephson junctions .
changing the bias - current gradually allows to follow the @xmath2-@xmath3-curves as in the current - biased experimental situation and to reproduce the hysteretic behaviour in particular .
we start with the discussion of the first branch of the @xmath2-@xmath3-curve , where one junction is in the resistive state .
quite generally the @xmath2-@xmath3-curve is expected to have peaks at the van hove singularities of the phonon dispersion .
details , however , depend on the oscillator strength defined by eq .
( [ osc ] ) which enters the longitudinal dielectric function eq .
( [ ephon ] ) .
in particular at the edge of the brillouin zone for @xmath192 only the motion of ions within the barrier contribute to the oscillator strength , the ions on the superconducting layers are inactive due to the factor @xmath193 in eq .
( [ zt ] ) .
these features will be illustrated in the following .
for the lattice dynamical model introduced above at @xmath192 only one type of particles is moving due to symmetry : in the acoustical branch the heavier ion , in the optical branch the lighter ion is moving . if the heavier ion is on the superconducting layers ( @xmath194 ) the oscillator strength vanishes at the end of the acoustic branch ( see fig .
[ fig3 ] ) , and peaks are expected to appear in the @xmath2-@xmath3-curve at the two van - hove singularities of the optical branch . on the other hand
, if the lighter ion is on the superconducting layers ( @xmath195 ) then the oscillator strength vanishes at the end of the optical branch , and peaks are expected at @xmath196 from the acoustical branch and at @xmath197 from the optical branch .
this is illustrated in fig .
[ fig4 ] where we have plotted results for the @xmath2-@xmath3-curve of the first branch in the two cases @xmath195 and @xmath198 . in our model
the phonon dispersion is fixed by the values of @xmath199 and the mass ratio @xmath200 . a measure for the oscillator strength
is the quantity @xmath201 . in fig .
[ fig4 ] we have used the following parameters : @xmath202 , @xmath203 and @xmath204 respectively .
the phonon frequencies are normalised to @xmath205 and are given by @xmath206 .
the oscillator strength is given by @xmath207 for @xmath195 and @xmath208 for @xmath198 .
the corresponding phonon dispersion and the oscillator strengths are shown in fig .
the values of the parameters @xmath52 , @xmath209 are adapted to tbcco , also the chosen value @xmath210 for the oscillator strength compares well to the experimental estimate @xmath211 @xcite .
the mass - ratio is chosen to obtain a sufficiently flat optical branch ( cf .
the band structure in ref . ) .
for @xmath195 the analytical results for the @xmath2-@xmath3-curve in fig.[fig4 ] show phonon peaks at the van hove singularity of the acoustic branch at @xmath192 and at the van hove singularity at @xmath197 of the optical branch , the resonance at the van hove singularity at @xmath196 of the optical branch is suppressed . for @xmath198
only structures due to the two van hove singularities of the optical branch appear . in both cases
the increase of the @xmath2-@xmath3-curve at low frequencies indicates the josephson plasma frequency .
the numerical results shown in fig . [ fig5 ] for @xmath195 ( here @xmath205 = 13.1 thz ) follow the analytical results in the regions of positive differential resistance perfectly , and otherwise show the hysteretic behaviour as seen in experiments . at low values of @xmath212
the @xmath2-@xmath3-curve switches back to the superconducing state of the junction .
another important effect of the coupling between josephson oscillations and phonons is the synchronisation ( phase - locking ) of josephson oscillations in different resistive junctions , which would be absent without phonons in short junctions , which are homogeneous parallel to the layers .
we want to illustrate this for the case of two resistive junctions coupled by one narrow optical phonon branch .
the second branch of the @xmath2-@xmath3-curve which corresponds to two resisitve junctions has a rather complex structure already for one phonon band .
this is shown schematically in fig .
if we denote the two dynamical states of the first branch by @xmath213 and @xmath214 , then on the second branch with two resistive junctions either both junctions are in state @xmath213 ( label @xmath215 ) , both junctions are in state @xmath214 ( label @xmath216 ) or one junction is in state @xmath213 while the other junction is in state @xmath214 ( label @xmath217 ) .
note that in the latter case the oscillation frequencies of the two junctions are different .
in the case of well separated resistive junctions the voltages of a structure for a given bias current are determined by @xmath218 , @xmath219 , @xmath220 .
this is no longer true , if the resistive junctions are close to each other and interact via phonons .
then the voltages in the second branch are slightly lower . in the case of two resistive junctions @xmath221 and @xmath27 two solutions of eq .
( [ multi1 ] ) exist with phase differences @xmath179 and @xmath222 , corresponding to in - phase and out - of - phase josephson oscillations , respectively . inserting these results in eq .
( [ multi1 ] ) two different @xmath2-@xmath3-curves can be calculated .
note that this formula applies only for the states @xmath215 and @xmath216 , because in the derivation we have assumed that the two junctions oscillate with the same frequency .
a similar formula can also be derived for the state @xmath217 . in that case
the phases are meaningless because the two junctions oscillate with different frequencies and are essentially decoupled . in fig .
[ fig7 ] we show examples calculated for a narrow optical band with @xmath223 and the light ion oscillating in the barrier . here
the different analytical solutions are shown together with numerical results for neighboring resistive junctions , @xmath224 ( a ) , and two resistive junctions separated by 1 or 2 superconducting junctions , @xmath225 ( b ) , @xmath226 ( c ) .
it is plausible that in the case of neighboring resistive junctions out - of - phase josephson oscillations ( @xmath227 ) favor a coupling to phonons at the edge @xmath228 of the brillouin zone , while the coupling of in - phase oscillations ( @xmath229 ) is strongest for zone - center phonons at @xmath197 .
this is shown in fig .
[ fig7]a , where the @xmath2-@xmath3-curve for the in - phase solution shows a peak at @xmath230 , while for the out - of - phase solution the current - maximum is at @xmath231 .
it can be seen that the numerical results in the dynamical state @xmath215 follow one of the analytical solutions before a hysteretic switch into state @xmath216 occurs ( outside the figure ) .
this is verified in fig .
[ fig7]a where the numerical @xmath2-@xmath3-curve follows the analytical curve for @xmath232 . in fig .
[ fig7]b the in - phase solution with @xmath233 has maxima for voltages corresponding to frequencies of optical phonons at both @xmath196 and @xmath197 , while the out - of - phase solution has a broad maximum in the middle of the brillouin zone .
the numerical results follow the in - phase solution . fig .
[ fig7]c shows results for @xmath234 .
the results obtained for two resistive junctions show that the peak position in the @xmath2-@xmath3-curve depends only slightly on the distance between the resistive junctions .
more important is the fact that phonons are able to synchronise the phases of josephson oscillations in different resistive junctions .
this is important for the use of such systems in high - frequency applications .
we have checked this numerically for the case of a block of many resistive junctions . for frequencies @xmath10 close to a phonon eigenfrequency @xmath231 at the edge of the brioullin zone
the josephson field oscillations in neighboring junctions differ by a value close to @xmath235 . for frequencies close to @xmath236 near the @xmath237-point the josephson oscillations are nearly in phase .
these phase - locked dynamical states are reached from arbitrary initial conditions for the phases .
there is also a synchronisation of josephson oscillations for frequencies far away from phonon resonances .
we did not yet investigate the stability of these states systematically but we expect these to be less stable than at frequencies close to a phonon resonance .
recently the explanation of the subgap resonances in refs . with the phonon coupling mechanism presented here could be well confirmed by raman measurements on the same samples @xcite and infrared reflectivity experiments with grazing incidence,@xcite where the latter allows to determine both longitudinal and transverse modes ( see table [ phonon_tab ] ) .
small deviations of the order of @xmath238 may be attributed to the fact that in optical experiments and in the intrinsic josephson effect different averages over @xmath239 of the dielectric functions are relevant .
note that in our theory modes which are raman active at @xmath197 may couple to intrinsic josephson oscillations also for @xmath240 .
earlier experimental data,@xcite which are obtained from polycrystalline samples , show the same qualitative behaviour , but differ in detail . with the help of rigid - ion @xcite and shell - model calculations @xcite some of the more pronounced structures
can be connected with certain elongation patterns of the ions in the unit cell .
for example , the peak in the @xmath2-@xmath3-curve at 4.64 thz in tl@xmath4ba@xmath4ca@xmath4cu@xmath5o@xmath241 seems to be due to a ( cu , ba)-mode .
the qualitative features of the subgap resonances have already been explained with the help of the phonon interpretation in ref .
: the position of the resonance is independent on temperature , magnetic field and the geometry of the probe , while the intensity of the structure varies @xmath242 with the critical current density @xmath243 .
also the behaviour of the position and intensity of the structures in external pressure are in agreement with the phonon explanation and formula ( [ iv - curve ] ) .
one of the main qualitative features of the general theory with dispersive phonon bands , which goes beyond the local oscillator model used in refs . , is the possibility to describe resonances at van - hove singularities , which appear e.g. at the upper band edge of the acoustical phonon band .
this might be an explanation for a resonance seen in ref . at @xmath244 mv ( @xmath245 ) in the @xmath2-@xmath3-characteristic of tl@xmath4ba@xmath4ca@xmath4cu@xmath5o@xmath241
, because the same frequency is expected by lattice dynamical calculations @xcite for the upper edge of the acoustical band , and there are no optical phonon bands in this low frequency range .
[ fig8 ] shows a fit of a @xmath2-@xmath3-curve calculated with the two - atomic chain model ( and some additional damping ) to the experimental results from ref . .
similarily in bscco the upper edge of the longitudinal acoustical phonon band , which has been detected in inelastic neutron scattering at @xmath246 mev @xcite , might correspond to a less pronounced phonon resonance in the intrinsic josephson effect at @xmath247 mv ( @xmath248 2.38 thz ) , which seems to be invisible in optical experiments . nevertheless at the present time this interpretation is not yet fully conclusive , as the instrumental resolution of @xmath249 mev in the neutron scattering experiment is still rather large compared with the accuracy in the measurement of the @xmath2-@xmath3-curve .
also the effect of the two van - hove singularities at the optical band edges on the @xmath2-@xmath3-curve as discussed above might have been seen in the satellite structures at 5.17 thz ( 10.7 mv ) and 5.6 ( 11.6 mv ) in the @xmath2-@xmath3-curve of bscco @xcite .
this interpretation is further supported by the theoretical prediction of the bandwidth @xmath250 and the fact that the assignment of no other phonon mode is plausible .
ccccccc + @xmath251 & 2.97 & 3.89 & 5.17 & 5.60 & ref . &
josephsoneffect + @xmath252 & 2.85 & & 5.07 & & ref . &
ir - reflecticity + @xmath252 & 2.86 & & 5.16 & & ref . &
ir - reflecticity + @xmath253 & & 3.80 & & & refs . &
raman effect + + @xmath251 & 3.63 & 4.64 & & & ref .
& josephsoneffect + @xmath252 & & 4.50 & & & ref . &
ir - reflecticity
in this paper we have developed a microscopic theory for the coupling between josephson oscillations and phonons in intrinsic josephson systems like the highly anisotropic cuprate superconductors .
we determined the precise form of the longitudinal dielectric function eq .
( [ ephon ] ) describing this coupling and obtained analytical results eqs .
( [ iv1],[multi1 ] ) for the @xmath2-@xmath3-curve for one and several resistive junctions .
the principle features and selection rules for phonon resonances in the @xmath2-@xmath3-curves are illustrated with help of a simple lattice dynamical model .
we have shown that not only optical , but also acoustical phonons at the edge of the brillouin zone couple to josephson oscillations .
this may explain the structure observed in ref . in the @xmath2-@xmath3-curve occurring at an unusual low voltage ( frequency ) , which is not found in reflectivity experiments , testing transverse optical phonons at @xmath254 , and in lattice - dynamical calculations for infrared active phonons at @xmath255 .
a weak satellite - structure observed in the @xmath2-@xmath3-curve of bscco @xcite may be due to a double resonance from the two van - hove singularites of an optical branch .
the analytical results are compared with results from a numerical integration of the coupled equations of motion for the josephson oscillations and phonons . for this purpose
a simplified lattice dynamical model has been used with one acoustical and one optical branch .
it is found that in the limit of large values of the mccumber parameter the numerical results follow closely the analytical solutions with the following exceptions : 1 . using a gradual change of the bias - current
, regions of the @xmath2-@xmath3-curve with negative differential conductivity are skipped as is observed in current - biased experiments .
2 . in the case of several resistive junctions , where several analytical solutions are obtained , the numerical result follows only one of the analytical solutions .
the stability of the different analytical solutions is currently investigated .
it seems to be that the solution which gives a minimum for the interaction energy between polarisation and the electric field generated by the josephson oscillations at a given frequency is most stable .
the phonons thus serve to synchronise the josephson oscillations in different resistive layers , which is important for the application of such systems as high - frequency devices . in this paper
we have considered only the case of a current distribution in the c - direction which is homogeneous along the layers .
we neglected all magnetic field effects assuming that all quantities are uniform along the layers . for real systems
this is an artificial approximation , because the current induced magnetic field can never be avoided completely .
nevertheless , we argue that this is a valid approximation for the intrinsic josephson systems forming a mesa structure of about fifty layers with a width of several @xmath256 and a thickness which is much smaller .
such a junction is still short with respect to the length @xmath257 describing the variation of the phase - difference along the layers ( in order to avoid completely the spontaneous formation of vortices , which have a smaller diameter , the mesa width should not exceed 1 - 2 @xmath256 ) .
therefore the treatment of the superconducting layers as metal sheets with a uniform charge distribution , the creation of uniform polarisation fields and the neglect of @xmath258 in the calculation of the longitudinal dielectric function is justified for the systems investigated .
this will be different in the case of longer junctions and strong external magnetic field @xcite .
in particular the flow of vortices and their interaction with phonons has to be investigated in this case @xcite .
the authors would like to thank a. yurgens and a. tsvetkov for discussions on their experimental results and for providing unpublished data , a. mayer and d. strauch for fruitful discussions on lattice dynamical aspects , and p. mller , r. kleiner , l. bulaevskii and a. bishop for the continuous support of our work on intrinsic josephson systems .
financial support by the deutsche forschungsgemeinschaft , the bayerische forschungsstiftung within the research project forsupra , the studienstiftung des deutschen volkes ( c.h . ) and the us department of energy under contract w-7405-eng-36 ( c.h . ) is gratefully acknowledged .
quite generally the rsj - equation for the n - th junction can be written as @xmath259 where @xmath260 is the dielectric function eq .
( [ ephon ] ) in real space and time domain .
let us denote the index of a general junction by @xmath261 and the subset of indices of resistive junctions by @xmath2 . in the following we assume that oscillations in all junctions are governed by one frequency @xmath10 and
the phases can be written as @xmath262 with @xmath263 expanding with respect to @xmath264 and keeping only the lowest harmonic one finds @xmath265 splitting the rsj - equation ( [ rsj2 ] ) into a dc - part and an oscillating part one obtains for the dc - part : @xmath266 while the oscillating part can be written as ( @xmath267 @xmath268 with @xmath269 writing @xmath270 the amplitude of the driving term for a resistive junction is given by @xmath271 from this an equation for @xmath275 ( @xmath173 ) in the resistive junctions is obtained : @xmath276 with @xmath277 using @xmath278 we get @xmath279^{-1 } e^{-i\theta_k},\ ] ] which gives us the dc - component eq .
( [ j ] ) of the current density @xmath280^{-1 } e^{-i\theta_k}. \end{aligned}\ ] ] finally we want to note that the dielectric function @xmath281 can also be used to write the rsj equations in a form where only the phases @xmath282 , of the resistive junctions @xmath173 enter : @xmath283 999 r. kleiner , f. steinmeyer , g. kunkel , p. mller , phys .
lett . * 68 * , 2394 ( 1992 ) .
r. kleiner , p. mller , phys . rev .
b * 49 * ( 1994 ) , 1327 .
k. schlenga , g. hechtfischer , r. kleiner , w. walkenhorst , p. mller , phys .
lett . * 76 * , 4943 ( 1996 ) .
a. yurgens , d. winkler , n. zavaritsky , t. claeson , proceedings of spie , vol * 2697 * , 433 ( 1996 ) .
p. seidel , a. pfuch , u. hbner , f. schmidl , h. schneidewind , t. ecke , j. scherbel , physica c * 293 * , 49 ( 1997 ) .
helm , ch .
preis , f. forsthofer , j. keller , k. schlenga , r. kleiner , p. mller , phys .
* 79 * , 737 ( 1997 ) .
helm , ch .
preis , f. forsthofer , j. keller , k. schlenga , r. kleiner , p. mller , physica c * 293 * , 60 ( 1997 ) . k. schlenga , r. kleiner , g. hechtfischer , m. mle , s. schmitt , p. mller , ch .
helm , ch .
preis , f. forsthofer , j. keller , h. l. johnson , m. veith , e. steinbei , phys .
b * 57 * , 14518 ( 1998 ) .
t. zetterer , m. franz , j. schtzmann , w. ose , h.h .
otto , k.f .
renk , phys .
rev , b 41 , 9499 ( 1990 ) ; v.m .
burlakov , s.v .
shulga , j. keller , k.f .
renk , physica c * 203 * , 68 ( 1992 ) and references therein .
buckley , m.p .
staines , d.m .
pooke , t. stoto , n.e .
flower , physica c * 248 * , 247 ( 1995 ) and references therein .
a. a. tsvetkov , d. duli , d. van der marel , a. damascelli , g. a. kaljushnaia , j. i. gorina , n. n. senturina , b. willemsen , n. n. kolesnikov , z. f. ren , j. h. wang , a. a. menovsky , t. t. m. palstra , phys .
b * 60 * , 13196 ( 1999 ) and references therein .
p. berberich , r. buemann , h. kinder , phys .
lett . * 49 * , 1500 ( 1982 ) .
ponomarev , e.b .
tsokur , m.v .
sudakova , s.n .
tchesnokov , m.a .
lorenz , m.a .
hein , g. mller , h. piel , b.a .
animov , solid state comm . * 111 * , 513 ( 1999 ) .
t. koyama , m. tachiki , phys .
b * 54 * , 16183 ( 1996 ) . ch .
preis , ch .
helm , j. keller , a. sergeev , r. kleiner , spie conference proceedings `` superconducting superlattices ii : native and artificial '' , vol * 3480 * , 236 ( 1998 ) .
e.g. maksimov , p.i .
arseev , n.s .
maslova , solid state comm .
* 111 * , 391 ( 1999 ) .
r. zeyher , g. zwicknagl , solid state commun .
66 , 617 ( 1988 ) ; r. zeyher , g. zwicknagl , z. phys .
b- condensed matter * 78 * , 175 ( 1990 ) .
a. d. kulkarni , j. prade , f. w. de wette , w. kress , u. schrder , phys .
b * 40 * , 2642 ( 1989 ) .
a. d. kulkarni , f. w. de wette , j. prade , u. schrder , w. kress , phys .
b * 43 * , 5451 ( 1991 ) . j. prade , a. d. kulkarni , f. w. de wette , u. schrder , w. kress , phys . rev .
b * 39 * , 2771 ( 1989 ) .
a. yurgens , private communication .
s. tajima , g. d. gu , s. miyamoto , a. odagawa , n. koshizuka , phys .
b * 48 * , 16164 ( 1993 ) .
jia , p .- y .
lin , y. xiao , x .- w .
jiang , x .- y .
gou , s. huo , h. li , q .- b .
yang , physica c * 268 * , 41 ( 1996 ) .
h. a. mook , b. c. chakoumakos , m. mostoller , a. t. boothroyd , d. mck .
paul , phys .
* 69 * , 2272 ( 1992 ) .
preis , ch .
helm , k. schmalzl , ch .
walter , j. keller , to appear in physica c , cond - mat/0001150 .
preis , k. schmalzl , ch .
helm , j. keller , in preparation . | in this publication a microscopic theory for the coupling of intrinsic josephson oscillations in layered superconductors with longitudinal c - axis - phonons is developed .
it is shown that the influence of lattice vibrations on the @xmath0-axis transport can be fully described by introducing an effective longitudinal dielectric function @xmath1 .
resonances in the @xmath2-@xmath3-characteristic appear at van hove singularities of both acoustical and optical longitudinal phonon branches .
this provides a natural explanation of the recently discovered subgap structures in the @xmath2-@xmath3-characteristic of highly anisotropic cuprate superconductors .
the effect of the phonon dispersion on the damping of these resonances and the coupling of josephson oscillations in different resistive junctions due to phonons are discussed in detail . |
mayall ( 1935 ) first recognized ic10 as an extragalactic object , and , in _ the realm of the nebula , _ hubble ( 1936 ) proposed that it was likely a member of the local group .
its location just 3 degrees out of the galactic plane has hampered investigations , but hubble s description of it as one of the most curious objects in the sky " has proven prophetic .
today ic10 is understood to be an irregular galaxy undergoing an intense burst of star formation likely triggered by infalling gas from an extended cloud which is counter - rotating with respect to the galaxy proper , as discussed by wilcots & miller ( 1998 ) , who conclude that ic10 is a galaxy that is still forming .
the starburst nature of ic10 was revealed primarily from the high number of wolf - rayet stars ( wrs ) found by massey , conti & armandroff ( 1992 ) and massey & armandroff ( 1995 ) .
hodge & lee ( 1990 ) had motivated these studies by their discovery of 144 h ii regions , the brightest of which were known to be comparable to the brightest seen in the smc ( hunter & gallagher 1985 ; kennicutt 1988 ) , a galaxy known to contain a substantial massive star population .
massey et al .
( 1992 ) used interference imaging to identify 22 wr candidates , of which 15 were confirmed spectroscopically ( massey & armandroff 1995 ) .
this number was quite unexpectly high .
ic 10 is about half the size of the smc ( van den bergh 2000 ) , which contains 11 wrs ( massey & duffy 2001 ) ; thus the overall surface density of wrs in ic10 is at least 5 times greater than in the smc .
the galaxy - wide surface density of wrs in ic10 is in fact comparable to that of the most active ob associations in m 33 ( massey & armandroff 1995 ) .
the distribution of wrs across ic10 shows that this high star - formation activity is not confined to a few regions ( which would simply be the result of statistical fluctuations or graininess " in the star formation rate ) , but rather is characteristic of a the galaxy as a whole .
this is the classic definition of a starburst galaxy ( hunter 1986 ; searle & sargent 1972 ) .
however , one of the very peculiar results of these wr studies was the abnormally large ratio of wc to wn stars given ic10 s metallicity ( log o / h+12=8.25 , skillman , kennicutt & hodge 1989 ; garnett 1990 ) . figure [ fig : wcwn ] shows the relative number of wc and wn stars plotted for different galaxies of the local group .
the interpretation of the strong trend with metallicity is straightforward : since the stellar winds of massive stars are driven by radiation pressure in highly ionized metal lines , stars of a given luminosity ( and mass ) will have a lower mass - loss rate in a lower metallicity system , and hence will lose less mass during their lifetime . in the conti scenario " ( conti 1976 , maeder & conti 1994 ) a massive star peels down like an onion due to mass - loss , revealing first the equilibrium products of the cno cycle at its surface ( wn stars ) , and then the he - burning products ( wc stars ) . very massive stars will therefore evolve first to the wn stage and subsequently to the wc stage , while a less massive star might evolve only through wn . at low metallicities
the wcs should come only from the very highest mass stars ( massey 2003 ) . the peculiar wc / wn ratio may be telling us something important about the star formation process in this , the nearest starburst galaxy .
if the imf was top - heavy ( or inverted ) , with an overabundance of the very highest mass stars , this could explain the results .
however , hunter ( 2001 ) finds a normal imf slope for the intermediate mass stars in ic10 .
it would be very odd for the imf of the highest mass stars to be decoupled from that of the intermediate - mass stars .
alternatively , if the burst that produced the wolf - rayet progenitors had been extremely coeval , then an abnormal wc / wn population could certainly result ( schaerer & vacca 1998 ) .
however , that would require a burst of duration less than 200,000 years over a scale of a kpc .
instead , a third , more prosaic possibility , is that the wc / wn ratio is strongly affected by incompleteness .
wn stars are much harder to detect than wc stars as their strongest emission lines are considerably weaker ( see massey & johnson 1998 ) .
massey & armandroff ( 1995 ) argued against this possibility on the basis that they had detected one wn star with lines as weak as commonly associated with wns .
however , since that time royer et al .
( 2001 , hereafter rsmv ) have reported the discovery of 13 new wr candidates in ic10 .
( one of these , their number 9 , is actually identical to star 6 in massey et al .
taken at face value , the additional stars would actually increase the wc / wn ratio rather than decrease it .
spectroscopy of this sample by crowther et al .
( 2002a ) confirmed 9 , leaving the wc / wn ratio little changed .
nevertheless , the study certainly calls into question the completeness of the original massey et al .
( 1992 ) sample .
the purpose of this _ letter _ is to report on the first results of a new , much deeper search for wolf - rayet stars in ic10 , along with spectroscopic confirmation of two of these stars .
complete details will be reported when the spectroscopic followup is complete .
however , the number of good wr candidates found is striking , and probably resolves the issue of the peculiar wc / wn ratio .
ic10 was imaged through three interference filters with the 4-m mayall telescope and mosaic ccd camera ( 8k x 8k ) .
the data were obtained on ut 19 and 20 sept 2001 .
the filter system is based upon that described by armandroff & massey ( 1985 ) and massey & johnson ( 1998 ) , but in large , 5.75 in by 5.75 in ( 146 mm @xmath3 146 mm ) format .
_ filter is centered at c iii @xmath4 , the strongest optical line in wc stars .
_ filter is centered on he ii @xmath5 , the strongest emission line in wn stars ( although it is also present in wcs ; see smith 1968 ) . the continuum filter _ ct _ is centered at @xmath6 .
the central wavelengths were designed for use in the fast , f/3 beam of the 4-m prime focus camera , and are roughly 50 in width .
the exposure time was 1.5 hrs in each filter , with the exposures broken into three 1800 sec exposures with the telescope dithered by 150 arcsecs ns and ew between exposures . the scale is 0.27 arcsec pixel@xmath7 , and the seeing on the nine images ranged from 0.85 to 0.97 arcsec , with an average of 0.92 arcsec .
the transparency was excellent .
( for comparison , the massey et al
. 1992 survey for wrs in ic10 used 1 hr exposures through each filter under 2.0 - 2.6 arcsec seeing with drifting clouds interrupting the exposures . )
exposures of spectrophotometric standards through the _ ct _ filter were used to determine the continuum magnitudes m@xmath8 , and are accurate to 0.1 mag .
the mosaic camera consists of 8 separate chips , each covering 18.4 arcmin ( ew ) by 9.2 arcmin ( ns ) , each large compared to the optical extent of ic10 ( half - light radius 2 arcmins ) .
ic10 had been centered on one of the chips , and it and a neighboring chip were reduced in the identical manner .
the latter was intended to serve as a control .
the data were processed through the iraf mosaic pipeline with refinements from the instrumental magnitudes were obtained using the point - spread - function fitting routine daophot " , as implemented under iraf . all together
, 114,000 stellar images were photometered . on average , 5300 stars
were measured on each frame of the control chip , and 7400 stars were measured on each frame of the galaxy chip .
candidate wrs were selected by comparing the magnitude differences _ wc - ct _ and _ wn - ct _ to the uncertainty in the magnitudes based upon photon statistics and read - noise , after a zero - point magnitude adjustment was made based upon the full ensemble of stars .
stars with magnitude differences more negative than @xmath9 mag and whose significance level was @xmath10 were considered valid candidates .
the 9 frames were treated as 3 independent sets , grouped by the three dithered telescope positions .
each candidate was examined on an image display by eye and checked for problems .
altogether the search revealed 238 unique candidates on the galaxy field , and 135 unique candidates for the control field ; many of the candidates were found multiple times .
we expect that none of the wr candidates " in the control field are real , given their location @xmath11 arcmin from ic10 .
such spurious detections are expected given the small magnitude differences we are looking for , and the possible presence of absorption features in the _ ct _ filter with non - wrs .
it is for this reason that we used a control field .
the search found _ all _ 24 of the spectroscopically confirmed wrs massey & armandroff 1995 ; crowther et al .
2002a ) with significance levels ranging from @xmath12 to @xmath13 . the weakest - lined star had a magnitude difference of @xmath14 mag , and the strongest - lined star @xmath15 mag between the wr filter ( _ wc _ or _ wn _ ) and the continuum filter _
ct_. p. crowther ( private communication ) kindly conveyed the specifics of which rsmv stars he had been able to confirm spectroscopically , and their spectral types in advance of publication .
we therefore note that our survey successfully distinguishes all of the known wcs from the known wns if we adopt a dividing line of _ wc - wn_=@xmath16 .
late - type wns will have strong n iii 4634,42 in the _ wc _ filter , while early - type wcs will have very broad c iii @xmath4 spilling over into the _
wn _ filter , so stars with a small absolute magnitude difference _
wc - wn _ are hard to classify just based upon our filter photometry .
the royer , vreux , & manfroid ( 1998 ) wr filter system uses five filters to help classify wr candidates to excitation subtype ( wn2 , wn3 ... wn9 ; wc4 , wc5 ... wc9 ) .
thus on the basis of their photometry alone , royer et al .
( 2001 ) announced the detection of very late - type wc ( wc9 ) stars in ic10 .
this result was highly surprising , as wc9 stars have previously been found only in much higher metallicity environments , for reasons thought to be well understood from stellar evolution : late - type wcs are thought to result from more enriched surface material , and the star can only peel down far enough to reveal these layers if the metallicity is high and mass - loss rates during the o - type stages are therefore high ( smith & maeder 1991 ) .
crowther et al .
( 2002b ) has recently called that into question , suggesting instead that the late - type wcs are the result of stronger stellar winds in the wr phase itself , but in either event wc9s are not expected in low metallicity environments .
this too was felt to be part of the puzzle of star formation in ic10 ( royer et al .
however , our survey detected _ none _ of these wc9 candidates .
an attempt to perform a quantitative spectroscopic analysis of these stars by crowther et al .
( 2002a ) using gemini failed to detect any emission .
we can probably conclude that these stars are not real wrs .
although the detection of all the known ic10 wrs , and the lack of detection of the spectroscopically rejected wr candidates gives us strong confidence in our survey , there is no substitute for spectroscopic confirmation . during a period of poor seeing at the mmt 6.5-m ( 14 sept 2002 ) we took time from our main program and observed two of our new candidates using a single slit setting . we used the 800 line / mm grating on the blue channel with a 2 arcsec slit to obtain 3.8 resolution spectra in the blue
; the exposure time was 2700 secs .
the spectra are shown in fig .
[ fig : spectra ] : both are wolf - rayet stars of wn type .
the coordinates are given in table 1 , along with the equivalent width ( ew ) and full - width - at - half - maximum ( fwhm ) of the he ii @xmath5 line .
the lines are sufficiently broad to rule out the possibility that either star is an of star , which might also show he ii and/or n iii emission .
ic10-wr24 is by far the brightest wr found in ic10 , and is likely a blend of a wr star and another star .
a blend would also explain the very weak emission ( table 1 ) combined with a normal line width . in figure
[ fig : compare ] we show a comparison of the photometry of wr candidates in the galaxy field with that from the control field . we expect that none of the 135 candidates in the latter are real .
given that both the galaxy field ( 238 candidates ) and control field covered an equal area , we calculate that ic10 may contain @xmath17 wolf - rayet stars in total .
although this number seems fantastically large , we note that 26 have now been confirmed spectroscopically : 15 by massey & armandroff ( 1995 ) , 9 by crowther et al .
( 2002a ) , and 2 here .
what then can we conclude about the statistics of wcs and wns in ic10 ?
first , for the spectroscopically confirmed wrs the wc / wn ratio is now 1.2 rather than 2.0 . if we simply take all of the wr candidates in the ic10 field , and correct by the number of wc " and wn " detections in the control field
, we would expect to find a ratio of 0.3 .
this ratio is only slightly higher than that of the outer region of m 33 , of similar metallicity ( figure [ fig : wcwn2 ] ) .
it is true that this result is somewhat dependent upon our choice of the dividing line between wc " and wn " in our photometry .
while our choice is consistent with our knowledge of the ic10 wr spectral types , spectroscopy of the remaining candidates will be needed to confirm this result .
if our statistical correction of the number of new candidates is correct , then spectroscopy should be able to confirm an aditional @xmath1 wcs and @xmath18 wns in ic10 .
even so , this may not represent the complete number , given the high reddening .
thus the mystery of the high wc / wn ratio in ic10 may be solved .
however , spectroscopic confirmation of such large additional number of wrs in ic10 would certainly make this galaxy even more unique in terms of its massive star population .
two pieces of evidence suggest that this may well be the case .
first , of the spectroscopically confirmed wns , two are of wn7 - 8 type . at low metallicities
the only similar late - type wns are found in the 30 dor region of the lmc , where very high mass stars abound .
studies of coeval regions containing these stars in the milky way suggest that they come from only the highest mass stars ( massey , degioia - eastwood , & waterhouse 2001 ) , and we would expect the progenitors to be even more massive in a low metallicity environment ( massey 2003 ) .
this is consistent then with ic10 having a normal imf but an exceptionally large population of massive stars .
secondly , the integrated h@xmath0 emission suggests that ic10 has one of the two highest rates of star formation per unit area known of a representative sample of non - interacting irregular galaxies ( hunter 1997 and private communication ) .
the active " area of ic10 is approximately 8 arcmin @xmath3 8 arcmin in angular extent ; at a distance of 660 kpc ( sakai , madore , & freedman 1999 ) , this correspond to an area of 2.2 kpc@xmath19 .
thus if our estimate is correct , ic10 would contain roughly 45 wolf - rayet stars kpc@xmath20 . for comparison
, the lmc contains @xmath1 wrs kpc@xmath20 ( massey & johnson 1998 ) .
a typical galactic ob association might contain several wrs , and be 100 pc in diameter , i.e. , with a surface density of a couple of hundred wrs kpc@xmath20 only several times larger than what we see _ globally _ in ic10 .
thus , if confirmed , the high number of wr stars would suggest that ic10 has a population of massive stars similar to that of an ob association but on a kpc scale .
crowther , p. a. , abbott , j. b. , drissen , l. , schild , h. , schmutz , w. , royer , p. , & smartt , s. j. 2002a in a massive star odyssey , from main sequence to supernova , iau symp .
k. a. vand der hucht , a. herrero , and c. esteban , ( san francisco : asp ) , in press | ic10 is the nearest starburst galaxy , as revealed both by its h@xmath0 surface brightness and the large number of wolf - rayet stars ( wrs ) per unit area .
the relative number of known wc- to wn - type wrs has been thought to be unusually high ( @xmath1 ) , unexpected for ic10 s metallicity . in this _
letter _ we report the first results of a new and deeper survey for wrs in ic10 .
we sucessfully detected all of the spectroscopically known wrs , and based upon comparisons with a neighboring control field , estimate that the total number of wrs in ic10 is about 100 .
we present spectroscopic confirmation of two of our wr candidates , both of which are of wn type .
our photometric survey predicts that the actual wc / wn ratio is @xmath2 .
this makes the wc / wn ratio of ic 10 consistent with that expected for its metallicity , but greatly increases the already unusually high number of wrs , resulting in a surface density that is about 20 times higher than in the lmc .
if the majority of these candidates are spectroscopically confirmed , ic10 must have an exceptional population of high mass stars . |
let @xmath5 be a 2-manifold .
a singular point @xmath6 of a @xmath7-map germ @xmath8 is a _
cuspidal edge _
if @xmath4 at @xmath9 is right - left equivalent to @xmath10 at the origin .
recently , the differential geometry of co - rank one singularities ( including cuspidal edges ) on surfaces was discussed by several geometers ( @xcite ) . in particular , in @xcite , isometric deformations of a special class of cross caps were discussed . relating to this ,
martins - saji @xcite defined several differential geometric invariants on cuspidal edges , and gave geometric meanings for them .
moreover , it was shown in @xcite that the limiting normal curvature @xmath3 defined in @xcite ( cf . )
is closely related to the behavior of gauss maps around cuspidal edges . on the other hand , the proof of the classical janet - cartan theorem on the local existence of isometric embeddings of real analytic riemannian @xmath11-manifolds into the euclidean space @xmath12 yields that any generic regular surface in @xmath0 has a non - trivial family of isometric deformations .
so it is natural to expect the existence of such non - trivial isometric deformations for surfaces with singularities . as shown in @xcite and @xcite , certain classes of ruled cross caps and cuspidal edges admit non - trivial isometric deformations , respectively . however , general cases have not been discussed yet .
along cuspidal edge singularities of a given @xmath7-map germ - maps , we consider only real analytic map germs in this paper since we shall apply the cauchy - kovalevskaya theorem.]@xmath13 , the unit normal vector field @xmath2 is well - defined as a smooth vector field of the surface .
let @xmath14 be a regular curve in @xmath5 satisfying @xmath15 as a parametrization of cuspidal edge singularities of the map @xmath4 .
we call @xmath14 the _ singular curve _ of @xmath4 .
we set @xmath16 which is a regular space curve .
let @xmath17 be the singular curvature function along the curve @xmath14 ( cf .
* ( 1.7 ) ) ) , and @xmath18 the limiting normal curvature along @xmath14 ( cf .
* ( 3.11 ) ) and ) .
then the curvature function of @xmath19 as a space curve is given by ( cf .
@xcite ) @xmath20 in @xcite , @xcite and @xcite , the singular curvature function @xmath17 and the limiting normal curvature function @xmath18 are considered as geometric invariants along cuspidal edge singularities , as well as the curvature function @xmath21 and the torsion function @xmath22 of @xmath19 .
the relationships amongst @xmath23 , @xmath3 and @xmath24 are given in @xcite .
an invariant @xmath25 of map germs at @xmath26 is called _ intrinsic _ if it is determined only by the first fundamental form ( cf .
the singular curvature @xmath23 is a typical intrinsic invariant of a cuspidal edge singularity . in @xcite ,
the cuspidal curvature @xmath27 at a given cuspidal edge singular point @xmath9 is defined .
let @xmath28 be the plane in @xmath0 passing through @xmath29 perpendicular to the vector @xmath30 .
then the intersection of the image of the singular set of @xmath4 by @xmath28 gives a @xmath31-cusp in the plane @xmath28 .
the value @xmath32 coincides with the cuspidal curvature of this @xmath31-cusp ( cf .
the following assertion holds : [ fact : prod ] the value @xmath33 is an intrinsic invariant .
( left ) and @xmath34 ( right),title="fig:",width=147 ] ( left ) and @xmath34 ( right),title="fig:",width=147 ] to prove our main result ( cf .
theorem [ thm : a ] ) , this fact plays a crucial role .
a cuspidal edge singular point @xmath9 is called _ generic _ if the osculating plane of @xmath35 is not orthogonal to @xmath2 at @xmath9 , that is , the limiting tangent plane does not coincide with the osculating plane of @xmath35 .
the gaussian curvature at a generic cuspidal edge is unbounded ( cf .
moreover , as shown in ( * ? ?
* theorem a ) , the following four conditions are equivalent : * a cuspidal edge singular point is generic ( cf .
figure [ fig : generic ] ) .
* the limiting normal curvature @xmath3 does not vanish at the singular point . *
the inequality @xmath36 holds ( cf . ) .
* let @xmath37 be the gaussian curvature and @xmath38 the signed area element of @xmath4 , where @xmath39 is a local coordinate system near the singular point @xmath9 .
then @xmath40 is well - defined on @xmath41 and does not vanish at @xmath9 .
we denote by @xmath42 the set of _ real analytic _ map germs of cuspidal edges @xmath43 which is defined on a neighborhood of the origin in @xmath44 .
moreover , let @xmath45 be the set of map germs of generic cuspidal edges . in this paper , we show the following : [ thm : a ] let @xmath17 be the singular curvature function of the singular curve @xmath14 of @xmath46 such that @xmath47 .
let @xmath48 be a real analytic regular space curve whose curvature function @xmath49 satisfies @xmath50 for all sufficiently small @xmath51 .
then there exist at most two map germs @xmath52 such that 1 .
the first fundamental form of @xmath53 coincides with that of @xmath4 , in particular , the singular curve @xmath14 in the domain of @xmath4 is the same as that of @xmath53 , 2 .
@xmath54 holds for each @xmath51 .
it is classically known that for a given planar curve @xmath14 having curvature function @xmath21 , there are at most two develpable surfaces having `` origami - singularities '' as a space curve whose curvature function @xmath55 satisfies @xmath56 ( see @xcite ) .
the above theorem can be considered as an analogue of this classical phenomenon .
kossowski @xcite is the first geometer who considered the realizing problem of given first fundamental forms as generic wave fronts .
however , in @xcite , the isometric deformations of singularities were not discussed , and the above theorem can be considered as a refinement of ( * ? ? ?
* theorem 1 ) in the case of cuspidal edge singularities .
since the intrinsic formulation of wave fronts are rather technical , the statement of theorem [ thm : a ] as a refinement of kossowski s realization theorem will be given and proved in the final section ( section [ sec : realize ] ) .
since kossowski applied the cauchy - kovalevskaya theorem to construct suitable second fundamental forms , our approach is completely different from his , and can be beneficial for the applications to isometric deformations .
however , it should be also remarked that kossowski s appoach will work for not only cuspidal edges but also other wave front singularities , ( for example , swallowtail singular points ) .
we get the following consequences of theorem [ thm : a ] : [ cor : b ] each map germ @xmath46 admits an isometric deformation @xmath57 in @xmath58 which moves the limiting normal curvature @xmath3 . in particular , @xmath3 and @xmath27 are not intrinsic invariants
is an extrinsic invariant , by construction the isometric deformation of ruled cuspidal edges satisfying @xmath59 . ] .
a given map germ @xmath60 is said to be _ planar _ ( resp .
_ non - planar _ ) if the image @xmath35 of the singular curve of @xmath4 is contained in a plane ( resp .
the torsion of @xmath35 does not equal to zero ) .
we show the following normalization theorem of generic cuspidal edges : [ cor : c ] for each @xmath46 , there exists a unique map germ @xmath52 of planar cuspidal edge singularities up to congruence such that * @xmath4 and @xmath53 induce the same first fundamental form , and * the curvature function of @xmath61 coincides with that of @xmath62 , where @xmath14 is the singular curve of @xmath4
. moreover , there exists a real analytic isometric deformation of @xmath4 into @xmath53 preserving the curvature function along the image of the cuspidal edge singularities . in the statement of corollary [ cor : c ]
, the condition @xmath63 can not be weakened to @xmath64 ( cf .
remark [ rem : non - generic ] and proposition [ prop : non - generic ] ) . as a consequence
, we also can not weaken the condition to @xmath65 in the statement of theorem [ thm : a ] .
two map germs @xmath66 are said to be _ congruent _ if there exist an ( orientation preserving or reversing ) isometry @xmath67 and a local analytic diffeomorphism @xmath68 such that @xmath69 . on the other hand , two map germs @xmath66 are said to be _ strongly isometric _ if there exist an isometry @xmath67 and a local analytic diffeomorphism @xmath68 satisfying the following properties : * @xmath70 and @xmath53 induce the same first fundamental form , and * the restriction of @xmath71 to the singular curve of @xmath4 coincides with that of @xmath53 .
a regular space curve @xmath72 passing through a point @xmath73 is called _ symmetric _ if there exists an isometry @xmath74 of @xmath0 which is not the identity map such that @xmath75 and the image of @xmath72 is invariant under the action of @xmath74 .
for example , the image of the singular curve of generic planar cuspidal edges are symmetric .
we get the following duality theorem for generic cuspidal edges : [ cor : d ] there exists an involution @xmath76 such that 1 .
@xmath77 is strongly isometric to @xmath4 , 2 .
if @xmath52 is strongly isometric to @xmath4 , then @xmath53 is congruent to @xmath4 or @xmath77 .
moreover , @xmath4 is not congruent to @xmath77 if the image of the singular set of @xmath4 is non - symmetric and non - planar .
we call @xmath77 the _ isomer _ of @xmath4 . for non - generic cuspidal edges
, the existence of isomers can not be expected in general ( see proposition [ prop : non - generic2 ] ) .
the proofs of these results are accomplished by an appropriate modification of the proof of the @xmath78-dimensional case of the janet - cartan theorem . in @xcite , we defined ` normal cross caps ' , expecting a similar normalization theorem as in corollary [ cor : c ] .
however , it seems difficult to apply the same technique , because cross cap singularities are isolated .
the fundamental tool to prove our results is the following fact ( cf .
@xcite ) : [ fact : ck ] let @xmath79 be a partial differential equation having @xmath80 @xmath81 as unknown functions , where @xmath82 is a real analytic map and @xmath83 this equation has a unique real analytic solution @xmath84 with an initial condition @xmath85 where @xmath86 @xmath87 are given real analytic functions .
the classical janet - cartan theorem for the existence of local isometric embeddings of real analytic riemannian @xmath78-manifolds can be proved as an application of this fact ( cf .
chapter 11 of @xcite ) .
our main results are also proved by applying fact [ fact : ck ] using the following special coordinate system along cuspidal edges : [ def : adapted ] let @xmath88 be a real analytic map and @xmath6 be a cuspidal edge singular point of @xmath4 .
( we can take a real analytic unit normal vector field @xmath2 defined on a neighborhood of @xmath9 . )
a real analytic local coordinate system @xmath89 at @xmath9 is called _ adapted _ if it satisfies the following properties along the @xmath90-axis : 1 .
[ item : adapted:1 ] @xmath91 , 2 .
[ item : adapted:2 ] @xmath92 , in particular , the singular set is contained in the @xmath90-axis , 3 .
[ item : adapted:3 ] @xmath93 is an orthonormal basis which is compatible with respect to the orientation of @xmath0 .
the existence of an adapted coordinate system was shown in ( * ? ? ?
* lemma 3.2 ) . throughout this paper
, we fix a real analytic map @xmath94 defined on a domain @xmath95 of the @xmath96-plane which has a generic cuspidal edge singular point at the origin @xmath97 , and assume that @xmath89 is an adapted coordinate system .
since @xmath89 is an adapted coordinate system , the _ limiting normal curvature _ @xmath3 of @xmath4 is given by ( cf .
equation ( 3.11 ) in @xcite ) @xmath98 where the dot @xmath99 is the inner product in @xmath0 . since @xmath92 along the @xmath90-axis
, there exists a real analytic map germ @xmath100 such that @xmath101 hold on a neighborhood of the @xmath90-axis . on the other hand ,
let @xmath102 be another real analytic map which has a generic cuspidal edge singular point at @xmath97 , and @xmath89 an adapted coordinate system .
similarly , there exists a real analytic map germ @xmath103 such that @xmath104 hold on a neighborhood of the @xmath90-axis .
[ prop : f ] let @xmath4 and @xmath53 be as above , and @xmath105 and @xmath103 as in and .
suppose that the first fundamental form of @xmath53 coincides with that of @xmath4 , then there exists a real analytic map @xmath106 such that @xmath107 where @xmath108 and @xmath109 @xmath57resp .
@xmath110 is the set of @xmath111-matrices @xmath57resp .
the set of regular @xmath111-matrices@xmath112 .
since @xmath4 and @xmath53 have the same first fundamental form on the same local coordinate system , we have @xmath113 that reduce to @xmath114 we define @xmath115 by @xmath116 ( { \varphi}\cdot f_{uu})_v -\dfrac{v}2{({\varphi}\cdot { \varphi})_{uu}}+v ( x\cdot x ) \end{pmatrix } } } , \end{aligned}\ ] ] where @xmath117 , @xmath118 and @xmath119 is the transpose of a regular matrix @xmath120 .
since @xmath121 it holds that @xmath122 the relation ( cf . and ) @xmath123 reduces to @xmath124 since ( cf . and )
@xmath125 it holds that @xmath126 on the other hand , by and , we have that @xmath127 since @xmath128 we have @xmath129 here , yields @xmath130 by and , we have @xmath131 the space curve @xmath132 parametrizes the image of the singular set of @xmath53 . since the cuspidal edge singularities of @xmath53 are generic , the osculating plane @xmath28 of the space curve @xmath133 is independent of the tangential direction @xmath134 of @xmath53 ( cf . ,
definition [ def : adapted ] ) .
since @xmath28 is spanned by @xmath135 , the matrix @xmath136 is regular .
hence , , and yield that @xmath137 in particular , @xmath138 attains the desired real analytic map .
let @xmath139 be as in the previous section .
then the space curve defined by @xmath140 gives a parametrization of the image of the singular curve @xmath141 of @xmath4 .
to prove theorem [ thm : a ] in the introduction , we prepare the following assertion : [ prop : ini ] let @xmath48 be a regular space curve satisfying such that @xmath51 is the arclength parameter .
then there exists a unique @xmath0-valued vector field @xmath142 @xmath57resp .
@xmath143 along @xmath72 satisfying the following properties : 1 .
[ item : w1 ] @xmath144 ( resp .
@xmath145 ) , 2 .
[ item : w2 ] @xmath146 ( resp .
@xmath147 ) , 3 .
[ item : w3 ] @xmath148 ( resp .
@xmath149 ) , where @xmath100 is given in , 4 . [
item : w4 ] @xmath150 ( resp .
@xmath151 ) .
since @xmath4 has non - zero limiting normal curvature , the curvature function @xmath21 of the space curve @xmath152 is positive ( cf . ) . since @xmath153 is a unit vector field ( cf .
definition [ def : adapted ] ) orthogonal to @xmath154 and @xmath155 , we have that @xmath156 let @xmath49 be the curvature function of @xmath48 . if we set @xmath157 the inequality yields @xmath158 on the other hand , since @xmath51 is the arclength parameter of @xmath72 , @xmath159 gives the principal unit normal vector field of the space curve @xmath48 . then applying the lemma in the appendix for @xmath160 , @xmath161 and @xmath162 as in , we can verify that @xmath163 ( resp .
@xmath164 ) satisfies [ item : w1][item : w4 ] .
the uniqueness of @xmath165 and @xmath166 also follows from the lemma in the appendix .
we consider the partial differential equation ( cf .
proposition [ prop : f ] ) @xmath167 for @xmath168 as in with the following initial conditions : @xmath169 by fact [ fact : ck ] , there exists a real analytic map @xmath170 on a sufficiently small neighborhood @xmath41 of the origin satisfying and .
since is equivalent to the conditions @xmath171 one can deduce the relation ( and also as a consequence ) directly from the relations and . consequently , the map @xmath172 ( resp .
@xmath173 ) has the same first fundamental form as @xmath4 and satisfies the properties ( 2 ) and ( 3 ) of theorem [ thm : a ] .
the ambiguity of the construction of @xmath53 as in the statement of theorem [ thm : a ] depends on the choice of the vector field @xmath174 along @xmath72 satisfying [ item : w1 ] , [ item : w2 ] and [ item : w3 ] . by the lemma in the appendix
, @xmath174 coincides with either @xmath175 or @xmath176 , which yields at most two possibilities for @xmath53 .
the proof of theorem [ thm : a ] is now reduced to the following proposition [ prop : ce ] . [
prop : ce ] the real analytic map @xmath177 has generic cuspidal edge singularities along the @xmath90-axis . since @xmath89 is an adapted coordinate system of @xmath53 , the vector field @xmath103 in is perpendicular to @xmath178 on the singular set .
moreover , by the definition of @xmath103 , @xmath179 is a unit normal vector field to @xmath53 which is well - defined on the singular set , where @xmath180 denotes the vector product in @xmath0 .
moreover , the function @xmath181 satisfies @xmath182 on the @xmath90-axis , because @xmath89 is an adapted coordinate system .
hence the singular points are _ non - degenerate _
* proposition 2.3 ) or ( * ? ? ?
* definition 1.1 ) ) .
moreover , since @xmath183 on the @xmath90-axis , the _
null direction _ at a point on the @xmath90-axis ( cf .
* page 306 ) or ( * ? ? ?
* page 495 ) ) is @xmath184 , which is linearly independent to the _ singular direction
_ @xmath185 .
thus , to show that @xmath53 is a cuspidal edge , it is sufficient to show that @xmath53 is a front ( cf .
@xcite or @xcite ) , which is equivalent to @xmath186 on the @xmath90-axis : let @xmath27 and @xmath3 be the cuspidal curvature ( cf .
* ( 2.4 ) ) ) and the limiting normal curvature of @xmath4 along @xmath187 , respectively .
since the singularity of @xmath4 consists of cuspidal edges , @xmath188 holds ( cf .
* lemma 2.8 ) ) , and since @xmath4 is generic ( i.e. the singularities of @xmath4 cossists only of generic cuspidal edges ) , @xmath189 .
by fact [ fact : prod ] , @xmath33 depends only on the first fundamental from .
thus , we have @xmath190 where @xmath191 and @xmath192 are the cuspidal curvature and the limiting normal curvature of @xmath53 , respectively . then by ( * ? ? ? * ( 2.4 ) ) , @xmath193 so it holds on the @xmath90-axis that @xmath194 hence the singular points of @xmath53 consist of cuspidal edge singularities .
moreover , by , the limiting normal curvature @xmath192 does not vanish , which implies that @xmath53 is generic .
since @xmath4 has non - zero limiting normal curvature , the curvature function @xmath21 of the space curve @xmath152 is greater than the absolute value @xmath195 of the singular curvature ( cf . ) .
let @xmath22 be the torsion function of @xmath19 .
for sufficiently small @xmath196 , there exists a regular space curve @xmath197 ( @xmath198 ) satisfying the following properties by the fundamental theorem of space curves * @xmath199 , * @xmath200 , * the curvature function of @xmath197 is equal to @xmath201 , * the torsion function of @xmath197 is equal to @xmath22 .
since @xmath202 is sufficiently small , we may assume that @xmath203 .
by theorem [ thm : a ] , there exists @xmath204 ( @xmath198 ) such that 1 .
the vector field @xmath205 satisfying @xmath206 is equal to @xmath207 along @xmath208 , where @xmath209 is a vector field along @xmath210 defined in proposition [ prop : ini ] , 2 .
the first fundamental form of @xmath211 is equal to that of @xmath4 , 3 .
the singular curve @xmath14 of @xmath4 is the same as that of @xmath211 , and 4 .
@xmath212 holds for each @xmath51 . since the geodesic curvature @xmath23 is intrinsic , yields that @xmath213 is equal to the absolute value of the limiting normal curvature of @xmath211 , which proves corollary [ cor : b ] .
let @xmath21 and @xmath22 be the curvature function and the torsion function of the space curve @xmath1 , respectively . for each @xmath214 $ ] , there exists a regular space curve @xmath197 satisfying the following properties by the fundamental theorem of space curves * @xmath199 , * @xmath200 , * the curvature function of @xmath197 is equal to @xmath21 , * the torsion function of @xmath197 is equal to @xmath215 .
since @xmath4 is generic , @xmath216 holds .
then by theorem [ thm : a ] , there exists @xmath217 ( resp .
@xmath218 ) for @xmath214 $ ] such that 1 . the vector field @xmath219 ( resp .
@xmath220 ) satisfying @xmath221 ( resp .
@xmath222 ) is equal to @xmath207 ( resp .
@xmath223 ) along @xmath208 , where @xmath224 are as in proposition [ prop : ini ] .
2 . the first fundamental form of @xmath225 ( resp .
@xmath226 ) is equal to that of @xmath4 , 3 .
the singular curve @xmath14 of @xmath4 is the same as that of @xmath225 ( resp .
@xmath226 ) , and 4 . @xmath227 ( resp .
@xmath228 ) holds for each @xmath51 . by this construction
, the torsion function of the curve @xmath229 vanishes identically .
so we can conclude that @xmath230 ( resp .
@xmath231 ) is a germ of a planar cuspidal edge . in particular
, @xmath229 lies in a plane @xmath28 , and the curve is invariant under the reflection with respect to the plane @xmath28 . let @xmath74 be the reflection with respect to the plane @xmath28 .
then we have @xmath232 in fact , the lemma in the appendix implies the fourth equality .
thus , by the uniqueness of fact [ fact : ck ] , we have @xmath233 .
this implies the uniqueness of @xmath53 as in corollary [ cor : c ] . by replacing @xmath2 with @xmath234
, we may assume that the limiting normal curvature @xmath3 of @xmath4 takes positive values without loss of generality .
by theorem [ thm : a ] , there exists @xmath235 ( resp .
@xmath236 ) for @xmath214 $ ] such that 1 . the vector field @xmath237 ( resp .
@xmath238 ) satisfying @xmath239 ( resp .
@xmath240 ) is equal to @xmath241 ( resp .
@xmath242 ) along @xmath208 , where @xmath243 are as in proposition [ prop : ini ] , 2 . the first fundamental form of @xmath172 ( resp .
@xmath244 ) is equal to that of @xmath4 , 3 .
the singular curve @xmath14 of @xmath4 is the same as that of @xmath172 ( resp .
@xmath244 ) , and 4 . @xmath245 ( resp .
@xmath246 ) holds for each @xmath51 .
since @xmath89 is an adapted coordinate system of @xmath4 , it holds that @xmath247 since @xmath248 and @xmath249 , we have @xmath250 so the uniqueness of @xmath251 implies that @xmath252 holds .
thus we have that @xmath253 we now define an involution @xmath254 it can be easily checked that @xmath255 is strongly isometric to @xmath256 . from now on ,
we suppose that the image of the singular curve of @xmath4 is non - symmetric and non - planar .
to prove corollary [ cor : d ] , it is sufficient to show that @xmath77 is not congruent to @xmath4 .
suppose that there exists an isometry @xmath74 in @xmath0 such that @xmath257 where @xmath258 is a local analytic diffeomorphism such that @xmath259 [ lem : xu ] under the situation above , we have @xmath260 where @xmath261 .
since @xmath74 is an isometry and @xmath89 ( resp .
@xmath262 ) is an adapted coordinate system for @xmath4 ( resp .
@xmath77 ) , the singular set @xmath263 of @xmath4 coincides with the singular set @xmath264 of @xmath77 .
hence we have @xmath265 let @xmath266 . since @xmath267 , @xmath268 here , we used the fact that @xmath262 is an adapted coordinate system for @xmath77
. hence we have @xmath269 ( @xmath261 ) .
since @xmath270 , we have the first conclusion . on the other hand , @xmath184 ( resp .
@xmath271 ) is the null direction of @xmath4 ( resp .
@xmath77 ) along the singular curve , so it holds that @xmath272 since @xmath273 on the singular set , we have @xmath274 .
since @xmath19 is non - planar , its torsion function does not vanish .
recall that the torsion function of a regular space curve does not depend on the choice of orientation of the curve , but changes sign by orientation reversing isometries of @xmath0 .
hence the isometry @xmath74 as in must be orientation preserving . since @xmath275 where @xmath276 , lemma [ lem : xu ] implies that @xmath277
. since @xmath267 and @xmath278 , it holds that , @xmath279 by definition of @xmath280 , it holds that @xmath281 so we can conclude that @xmath282 , and by lemma [ lem : xu ] , it holds that @xmath283 then @xmath51 is a common arclength parameter of @xmath19 and @xmath284 .
hence we have @xmath285 that is , the curve @xmath286 is symmetric at the origin ( @xmath287 ) , which contradicts our assumption . hence @xmath77 can not be congruent to @xmath4 . by the definition of strongly isometric equivalence , it is obvious that @xmath77 is strongly isometric to @xmath4 , and ( 2 ) of corollary [ cor : d ] holds .
[ rem : non - generic ] for a real analytic map germ @xmath4 of a non - generic cuspidal edge singularity , the partial differential equation can not be solved , since @xmath288 are not linearly independent .
cuspidal edges on surfaces of constant gaussian curvature are all non - generic ( cf .
moreover , as shown in proposition [ prop : non - generic ] below , theorem [ thm : a ] does not hold when @xmath4 is not generic .
although our method is not effective for such surfaces , examples of isometric deformations of flat cuspidal edges with vanishing @xmath3 are given in @xcite and @xcite .
[ prop : non - generic ] let @xmath14 be the singular curve of the cuspidal edge singularities of a @xmath7-map @xmath4 having vanishing gaussian curvature on its regular set .
suppose that @xmath289 lies in a plane in @xmath0 .
then the image of the curve @xmath35 lies in a straight line be a @xmath31-cusp on @xmath290-plane in @xmath0 . by considering a cylinder or a cone over @xmath291 , one can actually get a flat cuspidal edge whose image of singular set is contained in a line . ] .
the map @xmath4 is a flat front in the sense of @xcite . by (
* proposition 1.10 ) , the singular points of @xmath4 are not umbilical points
. then by ( * ? ? ?
* proposition 2.2 ) , @xmath4 is developable .
since @xmath3 vanishes along the singular curve @xmath187 , the asymptotic direction at @xmath1 is @xmath292 . in particular
, @xmath4 is a tangential developable surface , that is , we may set @xmath293 .
then the unit normal vector field @xmath2 of @xmath4 is equal to the binormal vector of @xmath294 .
suppose that @xmath1 is a regular curve with non - zero curvature function which lies in a plane .
since @xmath4 is a front , the fact @xmath295 implies that @xmath296 does not vanish , that is , the torsion function of @xmath35 does not vanish , which contradicts the fact that @xmath1 is a planar .
thus , the curvature function of @xmath1 vanishes identically , namely , its image lies in a straight line . in corollary
[ cor : d ] , we have shown the existence of isomers of given cuspidal edges . however , for the case of developable surfaces ( they are non - generic ) , there are no such isomers : [ prop : non - generic2 ] let @xmath48 be a regular space curve whose curvature function @xmath21 and torsion function @xmath22 have no zeros
. then there exists a unique flat front germ which has cuspidal edge singularities along @xmath72 .
let @xmath4 be a flat front which has cuspidal edge singularities along @xmath72 . by ( * ?
* proposition 1.10 ) , the singular set of @xmath4 can not be umbilical points .
so ( * ? ? ?
* proposition 2.2 ) implies that @xmath4 is a developable surface . in paticular
, @xmath4 must be a tangential developable of @xmath72 , which proves the assertion .
( bottom - left ) and @xmath297 ( top - right ) respectively , which have the common image of the singular set ( the range of the map is @xmath298 and @xmath299 ) . ]
finally , we shall give a concrete example : we set @xmath300 whose singular set consists of generic cuspidal edges , and is parametrized by @xmath301 we remark that the curvature function @xmath302 and torsion function @xmath24 of @xmath286 is given by @xmath303 in particular , @xmath35 is non - symmetric and non - planar as the singular curve of @xmath4 .
we set @xmath304 , where @xmath305 the singular set of @xmath53 consists of generic cuspidal edges , and coincides with that of @xmath4 , namely @xmath306 holds .
this @xmath53 gives an approximation of @xmath77 .
in fact , one can easily check that the coefficients of the first fundamental form of @xmath53 coincide with those of @xmath4 up to the fifth - order terms of @xmath307 near the origin ( see figure [ fig : isomer_ex ] ) .
let @xmath308 be a positive semi - definite _ real analytic _ metric defined on a neighborhood of the origin in the @xmath96-plane @xmath41 .
then @xmath308 can be written in the following form : @xmath309 the metric is called a ( germ of ) _ kossowski s metric _ ( cf .
@xcite ) if it satisfies the following conditions : 1 .
the @xmath90-axis consists of a singular set of @xmath308 , 2 .
@xmath310 along the @xmath90-axis , 3 .
there exists a real analytic function @xmath311 on @xmath41 such that @xmath312 , and 4 .
the gradient vector field @xmath313 does not vanish along the @xmath90-axis .
further systematic treatments of kossowski s metrics are given in @xcite .
let @xmath37 be the gaussian curvature of @xmath308 on @xmath314 .
kossowski showed in @xcite that @xmath315 can be smoothly extended on @xmath41 ( cf .
the condition ( d ) in the introduction ) , and proved the following assertion .
suppose that @xmath40 does not have zeros on the @xmath90-axis .
then there exist a neighborhood @xmath316 and a real analytic wave front @xmath317 such that the pull - back metric of the canonical metric of @xmath0 by @xmath4 coincides with @xmath308 .
see @xcite and @xcite for detailed discussions .
if the null - direction of the metric @xmath308 is transversal to the @xmath90-axis , the singular points of @xmath308 are called of _ @xmath318-singularities _ or
_ intrinsic cuspidal edges_. the induced metrics of wave fronts in @xmath0 are all considered as kossowski s metrics and cuspidal edge singular points corresponds to @xmath318-singularities ( cf .
moreover , in @xcite , the following expression of the singular curvature is given ; @xmath319 under the assumption that @xmath320 .
since this expression of @xmath23 does not depend on a choice of such a local coordinate @xmath89 , it can be considered as an invariant of the metric @xmath308 at @xmath318-singularities .
we can prove the following assertion as an modification of the proof of theorem [ thm : a ] : let @xmath308 be a real analytic kossowski s metric given as above and @xmath17 the singular curvature function along the @xmath90-axis .
let @xmath48 be a real analytic regular space curve whose curvature function @xmath49 satisfies @xmath321 for all sufficiently small @xmath51 .
suppose that @xmath40 does not have zeros on the @xmath90-axis .
then there exist a neighborhood @xmath316 and a real analytic wave front @xmath317 such that 1 . the first fundamental form of @xmath4 coincides with @xmath308 , 2 .
@xmath322 holds for each @xmath51 . in @xcite ,
the isometric deformations of singularities are not discussed , and theorem [ thm : a ] can be considered as a refinement of ( * ? ? ?
* theorem 1 ) in the case of cuspidal edge singularities .
as pointed out in the introduction , the following proof is a different from kossowski s original approach .
let @xmath323 be a kossowski s metric such that the singular set @xmath263 consists of @xmath318-singularities .
then , without loss of generality , we may assume that @xmath308 satisfies the following expressions ( cf .
@xcite ) ; @xmath324 where @xmath325 and @xmath326 are certain real analytic functions .
we now suppose that there exists a real analytic wave front @xmath327 so that the first fundamental form of @xmath4 is equal to @xmath308 .
we can define a real analytic map @xmath328 so that @xmath329
. then , it holds that @xmath330 . keeping in mind , we have that @xmath331 on the other hand , since @xmath332 we have @xmath333 .
in particular , we get ( cf . )
@xmath334 next in mind , we have that @xmath335 by , and , @xmath4 must satisfy the equation @xmath336 where @xmath337 , and @xmath338 is given by @xmath339 applying the cauchy - kovalevskaya theorem , we can get a real analytic solution of the equation under the same initial conditions as in the proof of theorem [ thm : a ] .
since the product curvature @xmath340 can be reformulated as an invariant of the @xmath318-singular point of a given kossowski metric as shown in @xcite , and the condition @xmath341 implies the condition @xmath342 .
thus we can prove that @xmath4 has cuspidal edge singularity along the @xmath90-axis , and we get the assertion .
the following assertion is applied to prove proposition [ prop : ini ] : let @xmath343 be the unit sphere in @xmath0 centered at the origin . let @xmath344 , @xmath345 be two mutually orthogonal unit vectors , and @xmath346 a real number with @xmath347 . then there exists a unit vector @xmath348 satisfying @xmath349 moreover ,
such a vector @xmath350 is uniquely determined under the assumption that the determinant @xmath351 is positive @xmath57resp .
negative@xmath112 .
let @xmath352 , where `` @xmath180 '' is the vector product of @xmath0 . then the vector @xmath353 has the desired property .
the uniqueness can be shown immediately .
the authors are grateful to wayne rossman for valuable comments .
m. kokubu , w. rossman , k. saji , m. umehara , and k. yamada , singularities of flat fronts in hyperbolic @xmath354-space , pacific j. math . , 221 ( 2005 ) , 303351 . l. f. martins and j. j. nuo - ballesteros , contact properties of surfaces in @xmath0 with corank @xmath356 singularities , preprint , 2012 .
available from www.uv.es/nuno/preprints/[email protected] .
l. f. martins and k. saji , geometric invariants of cuspidal edges , preprint , 2013 .
available from www.ibilce.unesp.br/home/departamentos/ matematica / singularidades / martins - saji - geometric.pdf r. oset sinha and f. tari , projections of surfaces in @xmath358 to @xmath359 and the geometry of their singular images , preprint , 2012 . available from www2.icmc.usp.br/@xmath355faridtari/papers/osettarisingularsurfaces.pdf .
m. spivak , a comprehensive introduction to differential geometry v , publish or perish inc .
houston , texas , 1999 .
k. saji , m. umehara , and k. yamada , the geometry of fronts , ann . of math . , 169 ( 2009 ) , 491529 . f. tari , on pairs of geometric foliations on a cross - cap , tohoku math .
j. 59 ( 2007 ) , 233258 . | along cuspidal edge singularities on a given surface in euclidean 3-space @xmath0 , which can be parametrized by a regular space curve @xmath1 , a unit normal vector field @xmath2 is well - defined as a smooth vector field of the surface .
a cuspidal edge singular point is called _ generic _ if the osculating plane of @xmath1 is not orthogonal to @xmath2 .
this genericity is equivalent to the condition that its limiting normal curvature @xmath3 takes a non - zero value . in this paper , we show that a given generic ( real analytic ) cuspidal edge @xmath4 can be isometrically deformed preserving @xmath3 into a cuspidal edge whose singular set lies in a plane . such a limiting cuspidal edge is uniquely determined from the initial germ of the cuspidal edge . |
different regression models have been proposed for lifetime data such as those based on the gamma , lognormal and weibull distributions .
these models typically provide a satisfactory fit in the middle portion of the data , but very often fail to deliver a good fit at the tails , where only a few observations are generally available .
the family of distributions proposed by birnbaum and saunders ( 1969 ) can also be used to model lifetime data and it is widely applicable to model failure times of fatiguing materials .
this family has the appealing feature of providing satisfactory tail fitting .
this family of distributions was originally obtained from a model for which failure follows from the development and growth of a dominant crack .
it was later derived by desmond ( 1985 ) using a biological model which followed from relaxing some of the assumptions originally made by birnbaum and saunders ( 1969 ) .
the random variable @xmath0 is said to be birnbaum
saunders distributed with parameters @xmath1 , say @xmath2-@xmath3 , if its cumulative distribution function ( cdf ) is given by @xmath4,\quad t > 0,\ ] ] where @xmath5 is the standard normal distribution function and @xmath6 and @xmath7 are shape and scale parameters , respectively .
it is easy to show that @xmath7 is the median of the distribution : @xmath8 . for any @xmath9 , then @xmath10-@xmath11 .
mccarter ( 1999 ) considered parameter estimation under type ii data censoring for the @xmath2-@xmath12 distribution .
lemonte et al .
( 2007 ) derived the second - order biases of the maximum likelihood estimates ( mles ) of @xmath6 and @xmath7 , and obtained a corrected likelihood ratio statistic for testing the parameter @xmath6 .
lemonte et al .
( 2008 ) proposed several bootstrap bias corrected estimates of @xmath6 and @xmath7 .
further details on the birnbaum saunders distribution can be found in johnson et al .
( 1995 ) .
rieck and nedelman ( 1991 ) proposed a log - linear regression model based on the birnbaum
saunders distribution .
they showed that if @xmath13-@xmath12 , then @xmath14 is sinh - normal distributed , say @xmath15 , with shape , location and scale parameters given by @xmath6 , @xmath16 and @xmath17 , respectively .
their model has been widely used as an alternative model to the gamma , lognormal and weibull regression models ; see rieck and nedelman ( 1991 , 7 ) .
diagnostic tools for the birnbaum saunders regression model were developed by galea et al .
( 2004 ) , leiva et al .
( 2007 ) and xie and wei ( 2007 ) , and the bayesian inference was considered by tisionas ( 2001 ) . in this paper
we propose a class of birnbaum saunders nonlinear regression models which generalizes the regression model introduced by rieck and nedelman ( 1991 ) .
we discuss maximum likelihood estimation of the regression parameters and obtain the fisher information matrix .
as is well known , however , the mles , although consistent , are typically biased in finite samples . in order to overcome this shortcoming , we derive a closed - form expression for the bias of the mle in these models which is used to define a bias corrected estimate . bias adjustment has been extensively studied in the statistical literature .
in fact , cook et al .
( 1986 ) proposed bias correction in normal nonlinear models .
young and bakir ( 1987 ) obtained bias corrected estimates for a generalized log - gamma regression model .
cordeiro and mccullagh ( 1991 ) gave general matrix formulae for bias correction in generalized linear models , whereas paula ( 1992 ) derive the second - order biases in exponential family nonlinear models .
cordeiro et al .
( 2000 ) obtained bias correction for symmetric nonlinear regression models .
more recently , vasconcellos and cribari neto ( 2005 ) calculate the biases of the mles in a new class of beta regression .
cordeiro and demtrio ( 2008 ) propose formulae for the second - order biases of the maximum quasi - likelihood estimates , whereas cordeiro and toyama ( 2008 ) derive the second - order biases in generalized nonlinear models with dispersion covariates .
the rest of the paper is as follows .
section [ reg_nonlinear ] introduces the class of birnbaum - saunders nonlinear regression models and discusses maximum likelihood estimation . using general results from cox and snell ( 1968 ) , we derive in section [ bias ] the second - order biases of the mles of the nonlinear parameters in our class of models and define bias corrected estimates
some special models are considered in section 4 .
simulation results are presented and discussed in section [ simulation ] for two nonlinear regression models .
we show that the bias corrected estimates are nearly unbiased with mean squared errors very close to the corresponding ones of the uncorrected estimates .
section [ application ] gives an application of the proposed regression model to a real fatigue data set , which provides a better fit at the tail of the data .
finally , section [ conclusions ] concludes the paper .
let @xmath13-@xmath3 .
the density function of @xmath18 has the form ( rieck and nedelman , 1991 ) @xmath19 this distribution has a number of interesting properties ( rieck , 1989 ) : ( i ) it is symmetric around the location parameter @xmath20 ; ( ii ) it is unimodal for @xmath21 and bimodal for @xmath22 ; ( iii ) the mean and variance of @xmath23 are @xmath24 and var@xmath25 , respectively .
there is no closed - form expression for @xmath26 , but rieck ( 1989 ) obtained asymptotic approximations for both small and large values of @xmath6 ; ( iv ) if @xmath27 , then @xmath28 converges in distribution to the standard normal distribution when @xmath29 .
we define the nonlinear regression model @xmath30 where @xmath31 is the logarithm of the @xmath32th observed lifetime , @xmath33 is an @xmath34 vector of known explanatory variables associated with the @xmath32th observable response @xmath31 , @xmath35 is a vector of unknown nonlinear parameters , and @xmath36 .
we assume a nonlinear structure for the location parameter @xmath37 in model ( [ eq1 ] ) , say @xmath38 , where @xmath39 is assumed to be a known and twice continuously differentiable function with respect to @xmath40 . for the linear regression @xmath41 , the model ( [ eq1 ] ) reduces to rieck and nedelman s ( 1991 ) model . the log - likelihood function for the vector parameter @xmath42 from a random sample @xmath43 obtained from ( [ eq1 ] ) , except for constants ,
can be expressed as @xmath44 where @xmath45/2)$ ] , @xmath46/2)$ ] for @xmath47 .
the function @xmath48 is assumed to be regular ( cox and hinkley , 1974 , ch .
9 ) with respect to all @xmath40 and @xmath6 derivatives up to third order .
further , the @xmath49 local matrix @xmath50 of partial derivatives of @xmath51 with respect to @xmath40 is assumed to be of full rank , i.e. , rank(@xmath52 for all @xmath40 .
the nonlinear predictors @xmath53 are embedded in an infinite sequence of @xmath54 vectors that must satisfy these regularity conditions for the asymptotics to be valid . under these assumptions ,
the mles have good asymptotic properties such as consistency , sufficiency and normality .
the derivatives with respect to the components of @xmath40 and @xmath6 are denoted by : @xmath55 , @xmath56 , @xmath57 , @xmath58 , @xmath59 , @xmath60 , etc .
further , we use the following notation for joint cumulants of log - likelihood derivatives : @xmath61 , @xmath62 , @xmath63 , etc .
let @xmath64 , etc .
all @xmath65 s and their derivatives are assumed to be of order @xmath66 . also , we adopt the notation @xmath67 and @xmath68 for the first and second partial derivatives of @xmath37 with respect to the elements of @xmath40 .
it is easy to see by differentiating ( [ eq2 ] ) that @xmath69 @xmath70 @xmath71 the score function for @xmath40 is @xmath72 , where @xmath73 is an @xmath74-vector whose @xmath32th element is equal to @xmath75 .
it is well - known that , under general regularity conditions ( cox and hinkley , 1974 , ch .
9 ) , the mles are consistent , asymptotically efficient and asymptotically normal .
let @xmath76 be the mle of @xmath42 .
we can write @xmath77 for @xmath74 large , where @xmath78 denotes approximately distributed , @xmath79 is the block - diagonal fisher information matrix given by @xmath80 , @xmath81 is its inverse , @xmath82 is the information matrix for @xmath40 and @xmath83 is the information for @xmath6 .
also , @xmath84 where @xmath85 is the error function given by @xmath86 details on @xmath87 can be found in gradshteyn and ryzhik ( 2007 ) . since @xmath79 is block - diagonal , the vector @xmath40 and the scalar @xmath6 are globally orthogonal ( cox and reid , 1987 ) and @xmath88 and @xmath89 are asymptotically independent .
it can be shown ( rieck , 1989 ) that @xmath90 for @xmath6 small and @xmath91 for @xmath6 large .
the mle @xmath92 satisfies @xmath93 equations @xmath94 for the components of @xmath40 and @xmath6 .
the fisher scoring method can be used to estimate @xmath40 and @xmath6 simultaneously by iteratively solving the equations @xmath95 where @xmath96 and @xmath97 for @xmath98 .
the above equations show that any software with a weighted linear regression routine can be used to calculate the mles of @xmath40 and @xmath6 iteratively . initial approximations @xmath99 and @xmath100 for the iterative algorithm are used to evaluate @xmath101 and @xmath102 from which these equations can be used to obtain the next estimates @xmath103 and @xmath104 .
these new values can update @xmath105 and @xmath106 and so the iterations continue until convergence is achieved .
we now obtain some joint cumulants of log - likelihood derivatives and their derivatives : @xmath107 @xmath108 @xmath109 let @xmath110 and @xmath111 be the @xmath112 biases of @xmath113 ( @xmath114 ) and @xmath89 , respectively .
the use of cox and snell s ( 1968 ) formula to obtain these biases is greatly simplified , since @xmath40 and @xmath6 are globally orthogonal and the cumulants corresponding to the parameters in @xmath40 are invariant under permutation of these parameters . from now on we use einstein summation convention with the indices varying over the corresponding parameters .
we have @xmath115 and @xmath116 where @xmath117 is the @xmath118th element of the inverse @xmath119 of the information matrix for @xmath40 , @xmath120 and @xmath121 denotes the summation over all combinations of parameters @xmath122 .
first , we consider equation ( [ vies - beta ] ) from which we readily have that the second sum is zero since @xmath123 .
it follows that @xmath124 by rearranging the summation terms we obtain @xmath125 let @xmath126 @xmath127 and @xmath128 @xmath129 be vectors containing the first and second partial derivatives of the mean @xmath37 with respect to the @xmath130 s .
we can write the above equation in matrix notation as @xmath131 where @xmath132 is the @xmath133th row of the @xmath134 identity matrix and vec@xmath135 is the operator which transforms a matrix into a vector by stacking the columns of the matrix one underneath the other .
it is straightforward to check that @xmath136 where @xmath137 and @xmath138 are @xmath49 and @xmath139 matrices of the first and second partial derivatives of the mean vector @xmath140 with respect to @xmath40 , respectively .
the @xmath112 bias vector @xmath141 of @xmath88 can then be written as @xmath142 where @xmath143 is an @xmath144 vector defined as @xmath145 .
we now calculate the @xmath112 bias of @xmath89 . using ( [ vies - alpha ] ) , we obtain @xmath146 where @xmath147 denotes the trace operator .
now , making use of the fact that @xmath148 , we can rewrite the @xmath112 bias of @xmath89 as @xmath149 equations ( [ bias - beta ] ) and ( [ bias - alpha ] ) represent the main results of the paper . the bias vector @xmath141 can be obtained from a simple ordinary least - squares regression of @xmath143 on the columns of @xmath150 .
it depends on the nonlinearity of the regression function @xmath151 and the parameter @xmath6 .
the bias vector @xmath141 will be small when @xmath143 is orthogonal to the columns of @xmath150 .
also , it can be large when @xmath152 and @xmath74 are both small .
equation ( [ bias - beta ] ) is easily handled algebraically for any type of nonlinear regression , since it involves simple operations on matrices and vectors .
for special models with closed - form information matrix for @xmath40 , it is possible to obtain closed - form expressions for @xmath141 . for linear models , the matrix @xmath153 and the vector @xmath143 vanish and hence @xmath154 , which is in agreement with the result due to rieck and nedelman ( 1991 , p. 54 ) that the mles are unbiased to order @xmath112 .
expression ( 6 ) depends directly on the nonlinear structure of the regression model only through the rank @xmath155 of @xmath150 .
it shows that the bias is always a linear function of the dimension @xmath155 of @xmath40 . in the right - hand sides of expressions ( [ bias - beta ] ) and ( [ bias - alpha ] ) , which are both of order @xmath112
, consistent estimates of the parameters @xmath40 and @xmath6 can be inserted to define bias corrected estimates @xmath156 and @xmath157 , where @xmath158 and @xmath159 are the values of @xmath141 and @xmath111 , respectively , at @xmath160 . the bias corrected estimates @xmath161 and
@xmath162 are expected to have better sampling properties than the classical mles @xmath88 and @xmath89 .
in fact , we present some simulations in section [ simulation ] to show that @xmath161 and @xmath162 have smaller biases than their corresponding uncorrected estimates , thus suggesting that these bias corrections have the effect of shrinking the adjusted estimates toward to the true parameter values .
however , we can not say that the bias corrected estimates offer always some improvement over the mles , since they can have mean squared errors larger .
it is worth emphasizing that there are other methods to obtain bias corrected estimates . in regular parametric problems , firth ( 1993 ) developed the so - called `` preventive '' method , which also allows for the removal of the second - order bias .
his method consists of modifying the original score function to remove the first - order term from the asymptotic bias of these estimates . in exponential families with canonical parameterizations ,
his correction scheme consists in penalizing the likelihood by the jeffreys invariant priors .
this is a preventive approach to bias adjustment which has its merits , but the connections between our results and his work are not pursued in this paper since they could be developed in future research .
additionally , it should be mentioned that it is possible to avoid cumbersome and tedious algebra on cumulant calculations by using efron s bootstrap ( efron and tibshirani , 1993 ) .
we use the analytical approach here since this leads to a nice formula .
moreover , the application of the analytical bias approximation seems to generally be the most feasible procedure to use and it continues to receive attention in the literature .
we now calculate the second - order bias @xmath163 of the mle @xmath164 of the @xmath32th mean @xmath38 .
we can easily show by taylor series expansion that @xmath165 where @xmath166 is a @xmath134 matrix of second partial derivatives @xmath167
( for @xmath168 ) , @xmath169 is the asymptotic covariance matrix of @xmath88 and the vectors @xmath170 and @xmath141 were mentioned previously .
all quantities in the above equation should be evaluated at @xmath88 .
the asymptotic variance of @xmath171 can also be expressed explicitly in terms of the covariance of @xmath88 by @xmath172
equation ( [ bias - beta ] ) is easily handled algebraically for any type of nonlinear model , since it involves simple operations on matrices and vectors .
this equation , in conjunction with a computer algebra system such as maple ( abell and braselton , 1994 ) will compute @xmath141 algebraically with minimal effort .
in particular , ( [ bias - beta ] ) may simplify considerably if the number of nonlinear parameters is small .
moreover , for any nonlinear special model , we can calculate the bias @xmath141 numerically via a software with numerical linear algebra facilities such as ox ( doornik , 2001 ) and r ( r development core team , 2008 ) .
first , we consider a nonlinear regression model which depends on a single nonlinear parameter .
equation ( [ bias - beta ] ) gives @xmath173 where @xmath174 and @xmath175 .
the constants @xmath176 and @xmath177 are evaluated at @xmath178 and @xmath89 to yield @xmath179 and the corrected estimate @xmath180 . for example , the simple exponential model @xmath181 leads to @xmath182 and @xmath183 . as a second example
, we consider a partially nonlinear regression model defined by @xmath184 where @xmath185 is a known @xmath186 matrix of full rank , @xmath187 is an @xmath144 vector , @xmath188 , @xmath189 and @xmath7 and @xmath190 are scalar parameters .
this class of models occurs very often in statistical modeling ; see cook et al .
( 1986 ) and cordeiro et al .
( 2000 ) . for example , @xmath191 ( gallant , 1975 ) , @xmath192 ( darby and ellis , 1976 ) and @xmath193 ( stone , 1980 ) .
ratkowsky ( 1983 , ch .
5 ) discusses several models of the form ( [ mspecial ] ) which include the asymptotic regression and weibull - type models given by @xmath194 and @xmath195 , respectively .
the @xmath49 local model matrix @xmath150 takes the form @xmath196 $ ] and , after some algebra , we can obtain from ( [ bias - beta ] ) @xmath197,\ ] ] where @xmath198 is a @xmath199 vector with a one in the last position and zeros elsewhere , @xmath200 ) is simply the set of coefficients from the ordinary regression of the vector @xmath201 on the matrix @xmath150 , and @xmath202 and @xmath203 are the large - sample second moments obtained from the appropriate elements of the asymptotic covariance matrix @xmath204 .
it is clear from ( [ b1 ] ) that @xmath141 does not depend explicitly on the linear parameters in @xmath205 and it is proportional to @xmath206 .
further , the covariance term @xmath203 contributes only to the bias of @xmath207 .
we now use monte carlo simulation to evaluate the finite - sample performance of the mles of the parameters and of their corrected versions in two nonlinear regression models .
the mles of the parameters were obtained by maximizing the log - likelihood function using the bfgs quasi - newton method with analytical derivatives .
this method is generally regarded as the best - performing nonlinear optimization method ( mittelhammer et al . , 2000 , p. 199 ) .
the covariate values were selected as random draws from the uniform @xmath208 distribution and for fixed @xmath74 those values were kept constant throughout the experiment . also , the number of monte carlo replications was 10,000 .
all simulations were performed using the ox matrix programming language ( doornik , 2001 ) . in order to analyze the performance of the estimates ,
we computed , for each sample size and for each estimate : the relative bias ( the relative bias of an estimate @xmath209 , defined as @xmath210 , is obtained by estimating @xmath211 by monte carlo ) and the root mean square error ( @xmath212 ) , where mse is the estimated mean square error from the 10,000 monte carlo replications .
first , we consider the nonlinear regression model @xmath213 where @xmath36 for @xmath47 .
the sample sizes were @xmath214 and 45 . without loss of generality ,
the true values of the regression parameters were taken as @xmath215 , @xmath216 , @xmath217 , @xmath218 and @xmath219 and @xmath220 .
table [ tab1 ] gives the relative biases of both uncorrected and corrected estimates to show that the bias corrected estimates are much closer to the true parameters than the unadjusted estimates .
for instance , when @xmath221 and @xmath222 , the average of the estimated relative biases for the estimates of the model parameters is @xmath223 , whereas the average of the estimated relative biases for the corrected estimates is @xmath224 .
hence , the average bias ( in absolute value ) of the mles is almost four times greater than the average bias of the corrected estimates
. this fact suggests that the second - order bias of the mles should not be ignored in samples of small to moderate size since they can be non - negligible .
the figures in table 2 show that the root mean squared errors of the uncorrected and corrected estimates are very close .
hence , the figures in both tables suggest that the corrected estimates have good properties . ccl rrrrr@xmath6 & @xmath74 & & @xmath225 & @xmath226 & @xmath227 & @xmath228 & @xmath229 + 0.5 & 15 & mle & 0.0006 & @xmath230 & 0.0011 & 0.0020 & @xmath231 + & & bce & 0.0007 & @xmath232 & 0.0001 & 0.0008 & @xmath233 + & 30 & mle & 0.0001 & @xmath230 & 0.0013 & 0.0009 & @xmath234 + & & bce & 0.0002 & @xmath235 & 0.0007 & @xmath236 & @xmath237 + & 45 & mle & 0.0003 & @xmath235 & 0.0007 & 0.0008 & @xmath238 + & & bce & 0.0003 & @xmath232 & 0.0003 & 0.0001 & @xmath239 + 1.5 & 15 & mle & @xmath240 & @xmath224 & 0.0248 & 0.0197 & @xmath241 + & & bce & @xmath242 & @xmath243 & 0.0113 & 0.0056 & @xmath244 + & 30 & mle & @xmath245 & @xmath246 & 0.0079 & 0.0078 & @xmath247 + & & bce & @xmath232 & @xmath248 & 0.0027 & 0.0012 & @xmath249 + & 45 & mle & @xmath250 & @xmath251 & 0.0052 & 0.0026 & @xmath252 + & & bce & @xmath253 & @xmath248 & 0.0023 & @xmath254 & @xmath255 + ccl rrrrr@xmath6 & @xmath74 & & @xmath225 & @xmath226 & @xmath227 & @xmath228 & @xmath229 + 0.5 & 15 & mle & 0.4093 & 0.4920 & 0.2707 & 0.0924 & 0.1234 + & & bce & 0.4093 & 0.4921 & 0.2709 & 0.0922 & 0.1067 + & 30 & mle & 0.3006 & 0.3806 & 0.2113 & 0.0688 & 0.0763 + & & bce & 0.3006 & 0.3806 & 0.2114 & 0.0686 & 0.0702 + & 45 & mle & 0.2434 & 0.2874 & 0.1768 & 0.0567 & 0.0590 + & & bce & 0.2434 & 0.2874 & 0.1769 & 0.0566 & 0.0555 + 1.5 & 15 & mle & 1.6302 & 1.1230 & 0.9756 & 0.3235 & 0.3938 + & & bce & 1.6333 & 1.1274 & 0.9819 & 0.3152 & 0.3315 + & 30 & mle & 0.9684 & 0.7003 & 0.5785 & 0.1931 & 0.2399 + & & bce & 0.9693 & 0.7011 & 0.5807 & 0.1908 & 0.2155 + & 45 & mle & 0.6505 & 0.5575 & 0.3895 & 0.1318 & 0.1837 + & & bce & 0.6507 & 0.5577 & 0.3901 & 0.1311 & 0.1700 + when the parameter @xmath6 increases , the finite - sample performance of the mles deteriorates ( see tables [ tab1 ] and [ tab2 ] ) .
for instance , when @xmath221 , the relative biases of @xmath207 ( mle ) and @xmath256 ( bce ) were 0.0020 and 0.0008 ( for @xmath219 ) and 0.0197 and 0.0056 ( for @xmath222 ) , which indicate an increase in the relative biases of nearly 10 and 7 times , respectively . also , the root mean squared errors in the same order were 0.0924 and 0.0922 ( for @xmath219 ) and 0.3235 and 0.3152 ( for @xmath222 ) .
next , we consider the very known michaelis menton model , which is very useful for estimating growth curves , where it is common for the response to approach an asymptote as the stimulus increases .
the michaelis
menton model ( ratkowsky , 1983 ) provides an hyperbolic form for @xmath37 against @xmath257 given by @xmath258 where the curve has an asymptote at @xmath259 . here
, the sample sizes were @xmath260 and 50 . also , the true values of the regression parameters were taken as @xmath261 and @xmath262 , with @xmath219 .
table [ tab3 ] gives the relative biases and root mean squared errors of the uncorrected and corrected estimates .
the figures in this table reveal that the mles of the parameters can be substantially biased , even when @xmath263 , and that the bias correction is very effective . in terms of mse ,
the adjusted estimates are slightly better than the ordinary mles .
cl rrr|rrr & & & + @xmath74 & & @xmath7 & @xmath190 & @xmath6 & @xmath7 & @xmath190 & @xmath6 + 20 & mle & 0.0476 & 0.1718 & @xmath264 & 0.6984 & 0.3947 & 0.0859 + & bce & @xmath245 & @xmath265 & @xmath266 & 0.5264 & 0.2783 & 0.0847 + 30 & mle & 0.0313 & 0.1077 & @xmath267 & 0.5245 & 0.2750 & 0.0684 + & bce & @xmath268 & @xmath269 & @xmath270 & 0.4478 & 0.2252 & 0.0678 + 40 & mle & 0.0215 & 0.0754 & @xmath271 & 0.4222 & 0.2207 & 0.0582 + & bce & @xmath236 & @xmath272 & @xmath273 & 0.3835 & 0.1954 & 0.0578 + 50 & mle & 0.0160 & 0.0558 & @xmath274 & 0.3609 & 0.1862 & 0.0516 + & bce & @xmath275 & @xmath236 & @xmath254 & 0.3380 & 0.1710 & 0.0514 +
obviously , due to the genesis of the birnbaum saunders distribution , the fatigue processes are by excellence ideally modeled by this model .
we now consider an application to a biaxial fatigue data set reported by rieck and nedelman ( 1991 ) on the life of a metal piece in cycles to failure .
the response @xmath276 is the number of cycles to failure and the explanatory variable @xmath277 is the work per cycle ( mj / m@xmath278 ) .
the data of forty six observations were taken from table 1 of galea et al .
( 2004 ) .
rieck and nedelman ( 1991 ) proposed the following model for the biaxial fatigue data : @xmath279 where @xmath280 and @xmath281 , for @xmath282 .
the mles ( the corresponding standard errors in parentheses ) are : @xmath283(0.3942 ) , @xmath284(0.1096 ) and @xmath285(0.0428 ) .
we take the logarithm of @xmath277 to ensure a linear relationship between the response variable ( @xmath286 ) and the covariate in ( [ rnmodel ] ) ; see galea et al .
( 2004 , figure 1 ) .
however , figure [ fig1 ] suggests a nonlinear relationship between the response variable and the covariate @xmath277 . here , we proposed the nonlinear regression model @xmath287 where @xmath36 .
the mles ( the standard errors in parentheses ) are : @xmath288(0.7454 ) , @xmath289(0.5075 ) , @xmath290(7.3778 ) and @xmath291(0.0417 ) .
the bias corrected estimates are : @xmath292(0.7734 ) , @xmath293(0.5266 ) , @xmath294(7.6548 ) and @xmath295(0.0433 ) .
hence , the uncorrected estimates are slightly different from the bias corrected estimates even for large samples ( @xmath296 observations ) .
figure [ fig2 ] gives the scatter - plot of the data , the fitted model ( [ nonlinear ] ) and the fitted straight line , say @xmath297 , where the mles are : @xmath298(0.1622 ) , @xmath299(0.0036 ) and @xmath300(0.0542 ) .
figure [ fig2 ] shows that the nonlinear model ( 10 ) ( unlike the linear model ) fits satisfactorily to the fatigue data . the @xmath301th observation ( the one with work per cycle near 100 )
can be an influential data .
however , it is not possible to say whether this observation is influential or not without using an efficient way to detect influential observations in the new class of models .
influence diagnostic analysis for this class of models will be developed in future research . following xie and wei ( 2007 )
, we obtain the residuals @xmath302 and @xmath303 .
figure [ fig3 ] gives the scatter - plot of @xmath304 versus the predicted values @xmath171 for both fitted models : ( i ) @xmath305 ; and ( ii ) @xmath306 .
figure [ fig3 ] shows that the distribution of @xmath304 is approximately normal for model ( ii ) but this is not true for model ( i ) .
based upon the fact that @xmath307 if @xmath308 , then the residual @xmath309 should follow approximately a sinh - normal distribution . versus @xmath171.,title="fig:",width=453,height=340 ] versus @xmath171.,title="fig:",width=453,height=340 ]
the birnbaum saunders distribution is widely used to model times to failure for materials subject to fatigue . the purpose of the paper was two fold .
first , we propose a new class of birnbaum saunders nonlinear regression models which generalizes the regression model described in rieck and nedelman ( 1991 ) .
second , we give simple formulae for calculating bias corrected maximum likelihood estimates of the parameters of these models .
the simulation results presented show that the bias correction derived is very effective , even when the sample size is large . indeed , the bias correction mechanism adopted yields adjusted maximum likelihood estimates which are nearly unbiased .
we also present an application to a real fatigue data set that illustrates the usefulness of the proposed model .
future research will be devoted to a study of diagnostics and influence analysis in the new class of nonlinear models .
we gratefully acknowledge grants from fapesp and cnpq ( brazil ) .
the authors are also grateful to an associate editor and two referees for helpful comments and suggestions . | we introduce , for the first time , a new class of birnbaum saunders nonlinear regression models potentially useful in lifetime data analysis .
the class generalizes the regression model described by rieck and nedelman [ 1991 , a log - linear model for the birnbaum saunders distribution , _ technometrics _ , * 33 * , 5160 ] .
we discuss maximum likelihood estimation for the parameters of the model , and derive closed - form expressions for the second - order biases of these estimates .
our formulae are easily computed as ordinary linear regressions and are then used to define bias corrected maximum likelihood estimates .
some simulation results show that the bias correction scheme yields nearly unbiased estimates without increasing the mean squared errors .
we also give an application to a real fatigue data set .
bias correction , birnbaum saunders distribution , maximum likelihood estimation , nonlinear regression . |
the spin - peierls ( sp ) transition@xcite in low dimensional magnetic compounds is the magnetic analog of the well known peierls transition@xcite in low dimensional metals .
the interest in this magneto - elastic phenomenon , which has been studied widely in organic compounds in the 1980s@xcite , has been strongly renewed with the discovery of inorganic materials ( @xcite , @xmath0-nav@xmath1o@xmath2 @xcite ) exhibiting this unusual phase transition .
in particular has attracted strong experimental and theoretical attention over the past few years@xcite .
there are at least three reasons for this : in the first place allows for the growth of large single crystals which permits for experiments which were difficult to perform in the past .
secondly , allows for well controlled substitutions ( e.g. , ) which provides a useful method to study its structural , and in particular its magnetic properties .
a surprising observation is that the sp phase transition is extremely sensitive to substitution ( 5 % of zn substitution for cu already makes the sp phase disappear ) .
equally surprising is that substitutions , which inherently increase the disorder in the system , lead to the formation of a long - range ordered antiferromagnetic phase at low temperatures@xcite . in the third place can not be considered as a `` classical '' spin - peierls system .
the susceptibility in the uniform phase ( t@xmath3 ) strongly deviates from the bonner - fisher behavior @xcite and the spin - peierls transition temperature strongly increases upon applying hydrostatic pressure despite an apparently decreasing @xmath4 exchange interaction@xcite .
there may be several reasons for the deviations from the classical behavior , notably a frustrating next nearest neighbor interaction@xcite , the two - dimensionality of the magnetism in the system@xcite , and the degeneracy of the energy scales for phonon and spin excitations in which calls for a description of in terms of mixed spin - phonon excitations rather than more or less decoupled spin and phonon excitations . at present , however , it is still unclear which of the above effects plays the dominant role in , although it has been shown that one can certainly not neglect the frustration in the system induced by the next nearest neighbor @xmath5 interactions .
the values reported for the frustration in , based on one - dimensional ( 1d ) adiabatic approximations , range from @xmath6@xcite , and may even be higher in under hydrostatic pressure@xcite . within the 1d adiabatic approximations
one then expects to have a spin - gap , even without dimerization , which is small for @xmath7 , but increases strongly upon approaching the majumdar - gosh point ( @xmath8)@xcite .
the influence of non - magnetic impurities on one - dimensional spin - peierls systems has been studied theoretically by several authors@xcite .
one of the main conclusions drawn from these studies is that one can not consider such an impurity as a local site defect .
in fact , it is expected that a non - magnetic impurity induces local antiferromagnetic correlations over a lengthscale which is determined by the spin - spin correlation length ( @xmath98 - 10 sites @xcite ) .
this enhancement of the local antiferromagnetic correlations may explain the extreme sensitivity of the sp phase to substitutions and possibly also @xcite the low temperature af long range order in and .
experimentally , evidence for these enhanced correlations may be found in raman scattering data on substituted compounds@xcite .
the raman phonon modes activated in the sp - phase of pure appear in substituted compounds already at much higher temperatures , and are even observed in compounds having a substitution level for which the sp - phase has completely vanished . pressure dependent experiments on @xcite have shown that strongly increases upon increasing pressure ( @xmath10 k / gpa ) , in good agreement with earlier predictions derived from zero pressure specific heat and thermal expansion experiments @xcite using the ehrenfest relation ( @xmath11 k / gpa ) .
it has been argued @xcite that the pressure dependence of may be due to a strong positive pressure coefficient of the @xmath12 frustration in ( @xmath13 ) .
recently , thermodynamic experiments on substituted compounds by t. lorenz _
_ @xcite have led to the surprising prediction that @xmath14 of substituted compounds is enhanced by a factor of @xmath15 over that of the pure compound .
motivated by these predictions , we report here on the pressure dependence of for a single crystal as determined from pressure dependent susceptibility and raman scatterig experiments . at low pressure , where the above mentioned prediction should be valid
, the susceptibility experiments indeed show a strongly enhanced @xmath14 , quantitatively in agreement with the predicted @xmath16 k / gpa @xcite .
in contrast to the pure compound , however , raman scattering experiments show that the pressure dependence of strongly deviates from a linear behavior , and approaches the pressure dependence of the pure compound for higher pressures .
the remainder of this paper is organized as follows : sections ii . and
describe the experimental results of the susceptibility and raman scattering experiments , respectively .
the subsequent section discusses these results in terms of a pressure ( and thus frustration ) dependent spin - spin correlation length .
finally , the last section presents the main conclusions of this paper .
the single crystal was grown from the melt by a floating zone method using an image furnace @xcite . for our measurements of the susceptibility under pressure we use a piece ( @xmath17mm@xmath18 ) of the crystal studied in ref .
@xcite . on the one hand the specific heat and thermal expansion data presented there
prove the high quality of this crystal .
on the other hand , using the same sample enables an unambiguous comparison between the measured pressure derivatives of t@xmath19 and the predictions from thermodynamic properties @xcite .
the ac susceptibility was measured in a pressure range @xmath20 gpa with a mutual inductance bridge at a frequency of 6.1 khz in a magnetic field of about 1 oe . both , the primary coil and the astatic secondary coils are located inside the teflon cup of a piston cylinder pressure cell which is similar to that described by j.d .
thompson @xcite .
a methanol : ethanol ( 4:1 ) mixture was used as pressure transmission medium in order to guarantee hydrostatic pressure conditions .
the pressure was controlled at low temperatures by simultaneously measuring the superconducting transition temperature of a small piece of lead . in fig .
1 the raw data of the ac susceptibility measurements are presented for several pressures .
the overall temperature dependence of the ac signal is mainly determined by a smooth background , which does not change strongly as a function of pressure .
this background signal is nearly temperature independent above about 12 k and increases at lower temperatures @xcite . due to this background and its pressure dependence
a precise quantitative extraction of the susceptibility of is not possible . in particular
, one can not separate the curie - like increase of @xmath21 present in doped at low temperatures @xcite from the background signal . however , the pronounced anomalies at the spin - peierls transition , which is associated with a rapid decrease of the susceptibility in a small temperature range , are clearly visible at all pressures .
besides the apparent increase of t@xmath19 we do not find significant pressure dependencies of the susceptibility anomaly at the corresponding t@xmath19 .
as shown e.g. in ref .
@xcite the spin - peierls transition temperature can be derived precisely from a sharp maximum of the temperature derivative of the magnetic susceptibility .
the t@xmath19 obtained in this way are shown as a function of pressure in fig .
together with the data we show a line , which represents a pressure derivative of @xmath22k / gpa as predicted from the ehrenfest relation for @xmath23 @xcite .
the measured , nearly linear , pressure dependence of up to 0.43 gpa is in excellent agreement with the predicted behavior .
fitting the data obtained from the susceptibility measurements by a linear increase of t@xmath24 reveals an only slightly smaller slope of of @xmath25k / gpa .
we mention , however , that this slope systematically increases with decreasing pressure range considered for the linear fit .
for example , fitting the data up to 0.27 gpa yields 8.2(5)k / gpa , whereas the slope @xmath26 above 0.2 gpa only amounts to 6.3(7)k / gpa .
thus the susceptibility data are also consistent with a slight decrease of @xmath26 with increasing pressure .
the susceptibility experiments are particularly suited for the low pressure region ( @xmath27 gpa ) , where also the prediction made in ref.@xcite should hold .
these low pressure data are shown in fig . 2 ( filled symbols ) which also displays a linear fit to the data ( solid line ) and the prediction made in ref.@xcite ( dashed line ) .
the measured , nearly linear pressure dependence of @xmath28 clearly agrees well with the predicted behavior in this pressure range . comparing the data on presented so far to the results for pure
it is apparent that the pressure derivative in the doped compound is significantly larger . in other words ,
the reduction of @xmath29 due to doping with 1.4% zn decreases from @xmath30k at @xmath31 to @xmath32k at @xmath33 gpa . since an extrapolation of our data of doped would yield the unreasonable result that @xmath29 increases above the value of pure at high pressures ( @xmath34gpa ) , we have studied the behavior in this pressure range using raman scattering
pressure and temperature dependent raman scattering experiments have been performed on thin platelets ( @xmath35 @xmath36m@xmath18 ) of the same single crystal as used for the susceptibility measurements .
these platelets were mounted in a clamp - type diamond anvil cell , using a methanol / ethanol mixture as the pressure medium .
the cell has subsequently been mounted in a he flow cryostat with a temperature regulation better than 1k .
polarized raman spectra have been recorded in a quasi - backscattering geometry using the 514-nm line of an ar@xmath37-laser for excitation and a ccd equipped @xmath38 spectrometer for detection .
the polarization of incident and scattered light have both been chosen parallel to the c - axis of the orthorhombic crystal structure ( _ i.e. _ along the cu - o chains ) .
typical spectra observed for different pressures at @xmath39 k are presented in fig .
3 . for atmospheric pressure one observes three strong peaks at 184 , 330 and 592 @xmath40 which have been assigned to a@xmath41-modes of the orthorhombic crystal structure @xcite .
the four additional modes observed at 26 , 105 , 226 , and 370 are characteristic of the sp phase @xcite .
the two modes at 105 and 370 can be assigned to brillouin zone boundary phonons , and are activated in the sp - phase through the magnetic and/or structural cell doubling .
the low frequency mode at 26 arises due to transitions from the singlet ground state to the lowest magnetic singlet state@xcite .
the energy of this mode is somewhat smaller than that of the corresponding mode in the pure compound ( 30 ) , indicating that zn - substituted crystals have a smaller spin - gap .
this is in good agreement with the observed reduction of the transition temperature in ( @xmath28[]@xmath42 11 k ; @xmath28[]@xmath42 14 k ) . finally , the broad , asymmetric response , peaking at 226 is generally thought to arise from two - magnon scattering due to the fleury - loudon @xcite mechanism , and should reflect a weighted two - magnon density of states .
similar to the situation in the pure compound@xcite , the application of hydrostatic pressure leads to drastic changes in the observed raman spectra . with increasing pressure the two - magnon peak shifts to lower frequencies ( 208 @xmath40 at 2.9 gpa ) , reflecting an increasing magnetic frustration in the system .
in contrast to this , but for the same reasons , the gap - related mode at 26 @xmath40 rapidly shifts to higher frequecies upon applying pressure .
simultaneously this mode also broadens , and at 2.9 gpa the singlet response is hardly observed anymore .
most likely , this broadening is due to non - hydrostatic effects at higher pressure , as is also reflected in the width of the phonon modes .
a good method to determine @xmath28 for the pure compound is to monitor the temperature dependence of the singlet mode@xcite .
as it is the case for the triplet spin - peierls gap@xcite , the energy of the singlet mode strongly decreases upon increasing temperature until it vanishes at @xmath28 .
alternatively , it has been shown that also the intensity of the spin - peierls active 370 phonon may be used to determine @xmath28 @xcite .
the intensity of this mode shows a temperature behavior characteristic for a second order phase transition , _
i.e. _ an intensity which varies as the square of the order parameter .
this is demonstrated for in fig .
4a , which shows a comparison between the temperature dependence of the intensity of the 370-phonon ( filled symbols ) with that of the the experimentally determined spontaneous strain ( solid line)@xcite , which should vary as the square of the structural order parameter . for the pure compound
the above mentioned methods yield essentially the same results and are in good agreement with the transition temperature obtained using other methods ( e.g. , thermal expansion , susceptibility , heat capacity ) , as shown in refs .
@xcite .
since the present data do not allow to use the singlet response as an indicator for the presence of the sp phase over the whole pressure range , we will use the intensity of the 370 mode to determine @xmath28 . as an example , the inset of fig .
4b shows the temperature dependence of this mode for @xmath43 gpa .
relative to the @xmath44 mode at 347 , this mode , which is observed as a relatively strong lorentzian shaped mode at @xmath39 k , rapidly looses intensity as the temperature increases , and finally disappears around @xmath45 k. figure 4b shows the temperature dependence of the intensity of the 370-phonon as function of temperature ( normalized to the intensity at the lowest temperature ) for @xmath46 and @xmath47 gpa . for higher pressure ( @xmath48 gpa )
the intensity indeed shows the expected behavior , _
i.e. _ it qualitatively seems to reflect the squared order parameter of a second order phase transition .
in contrast to this , and to the situation in the pure compound , the temperature dependence of the intensity of the 370 mode at @xmath49 does not seem to reflect the square of the structural order parameter .
this becomes more clear from a comparison of the temperature dependence of the intensity of the 370- mode with the spontaneous strain measured on the same sample ( fig .
4b , solid line ) .
although the spontaneous strain still shows the behavior of a relatively well defined second order phase transition with @xmath50 k , the intensity of the 370 mode drops much more rapidly with temperature , and , moreover , remains non - zero up to temperatures well above the phase transition @xcite .
these observations put question marks to the interpretation of a structurally induced intensity for the 370- mode in the sp phase .
indeed , the intensity of this mode in the pure compound also shows a good correlation with the square of the singlet energy ( the square of the `` magnetic order parameter '' ) .
this is not surprising since the singlet energy directly correlates with the sp gap @xmath51 , which itself is related to the structural order parameter @xmath52 ( @xmath53 within cross - fisher theory @xcite ) .
this indicates that the intensity of the 370-mode may not be induced by the structural distortion in the sp phase , but rather by the doubling of the magnetic unit cell @xcite .
it furthermore strongly suggests that the relation between the structural distortion and the size of the sp gap may drastically change upon substitutions , strongly deviating from the cross - fisher prediction @xcite .
at low pressures the transition temperature in the substituted compound indeed increases much faster with pressure ( @xmath54 k / gpa ) as compared to the increase for the pure compound ( @xmath55 k / gpa ) ( see fig .
2 ) . the steep linear behavior observed for low pressures clearly can not persist indefinitely .
indeed , the transition temperatures obtained for higher pressure from the raman data show a pronounced non - linear behavior at higher pressures ( see fig . 5 , full squares ) .
in this regime , the pressure dependence of the transition temperature for @xmath56 approaches that of the pure compound . in order to get some understanding of the strong pressure dependence of @xmath28 in , consider first the strong dependence of @xmath28 on substitutions . as has been shown by , e. g. , fukuyama _ et al . _
@xcite , a substitution at a single site does not only perturb this site , but leads to the formation of local af ordering and suppression of the dimerization over several lattice sites .
the spatial extent of the impurity is typically of the order of the spin - spin correlation length @xmath57 .
this immediately explains the strong reduction of @xmath28 upon substitution , and may possibly also explain the occurrence of a long - range ordered af ground state in the substituted crystals@xcite .
as has been pointed out by mostovoy and khomskii@xcite , one expects the sp transition to be fully suppressed when the average distance between the impurities becomes of the order of the spin - spin correlation length .
_ i.e. _ for concentrations of about 10 % ( @xmath58 )
. the strong increase of @xmath28 with increasing pressure in pure @xmath59 has been discussed in terms of a strongly increasing frustration @xcite .
this may , of course , also explain part of the observed increase of @xmath28 in the substituted compound .
there is , however , a second effect which arises from the increasing frustration and is of importance in the substituted compounds .
as the frustration increases , one expects the spin - spin correlation length to decrease .
indeed , at @xmath8 ( the mg point ) one expects the ground state to be formed by singlets pairs and to have a correlation length of exactly two spins .
numerical density matrix renormalization group calculations of the correlation length have indeed confirmed this picture @xcite .
the consequence of the decreasing correlation length is that the influence of impurities is strongly suppressed upon applying pressure .
this actually may explain the observed behavior of @xmath60 in .
initially , the increase of the transition temperature is driven by a combination of an increasing frustration and a decreasing influence of the impurities with frustration . at higher pressure
the effect of the impurities is strongly suppressed , and the increase of @xmath28 is fully driven by the increasing frustration alone . at these pressures
one therefore expects only small differences between the pure and the substituted compounds , in good agreement with the observed behavior .
a further indication of the suppression of the effects of the impurities may be found from the temperature dependence of the intensity of the 370-mode ( see fig .
as mentioned in the previous section , at low pressures the behavior of the intensity of this mode does not agree well with the temperature dependence of the structural order parameter .
in addition , the mode remains observable at temperatures well above the phase transition .
this latter observation is consistent with the idea that the mode is induced by magnetoelastic interactions and the magnetic cell doubling .
if indeed the presence of impurities leads to enhanced af correlations around the impurities , then this also leads to a local magnetic unit cell doubling , and hence to an activation of the 370- mode in the raman spectra .
of course , at high enough temperatures the intensity should still disappear due to the decreasing correlation length , _ i.e. _ the decreasing spatial `` size '' of the impurities . at high enough pressure ( @xmath48 gpa )
the temperature dependence of the intensity of the 370- mode is again in qualitative agreement with the dependence observed for the pure compound , in good agreement with the idea that the influence of the impurities is strongly reduced for high pressures .
in this paper we have shown , by two different methods , that the pressure dependence of the spin - peierls transition temperature in zn - substituted is strongly enhanced over that of the pure compound .
the initial - slope data are found to be in excellent agreement with the predictions using the ehrenfest relation@xcite . at higher pressure the pressure coefficient @xmath14 of approaches that of the pure compound . the strong enhancement at low pressure , and the approach to the behavior of the pure compound at high pressure is in good agreement with an increasing frustration in these compounds , which leads to a decreasing correlation length and thus to a decrease of the influence of the impurities .
we gratefully acknowledge g. bouzerar for many fruitful discussions , and for the calculations on the frustration dependence of the correlation length .
this work was supported by dfg through sfb 341 and by bmbf 13n6586/8 .
laboratoire de chimie des solids is unit de recherche associe cnrs n@xmath61 446 .
fig . 1 : raw data of the ac susceptibility measurements ( output signals of the lock - in amplifier ) for several pressures .
the data are subsequently shifted by @xmath62 v for clarity .
2 : pressure dependence of @xmath28 as determined from the susceptibility data of @xmath63 ( filled symbols ) .
the lines gives the predicted pressure dependencies from the ehrenfest relation ( solid line : @xmath64 k / gpa ( ref .
@xcite ) . dashed line : @xmath65 k / gpa @xcite . )
fig . 3 : ( zz ) polarized raman spectra of recorded at t=7 k for @xmath66 and @xmath67 gpa .
the curves have been given an offset for clarity .
the positions of observable ar@xmath37-laser plasma lines are denoted by an asterisk
. + fig . 4 : ( a ) intensity 370-mode in @xmath59 as a function of temperature for @xmath49 ( filled symbols ) , and @xmath68 gpa ( open symbols ) .
the solid line gives the experimental temperature dependence of the spontaneous strain at @xmath49 gpa @xcite .
dashes line is a guide to the eye .
( b ) intensity of the 370- mode in @xmath56 as function of temperature for @xmath49 ( filled symbols ) , @xmath68 and @xmath47 gpa ( open symbols ) .
the solid line gives the experimental temperature dependence of the spontaneous strain at @xmath49 gpa @xcite .
dashes lines are a guide to the eye .
the inset shows the temperature dependence of the ( zz ) polarized spectrum of @xmath56 for @xmath69 gpa .
+ fig . 5 : dependence of as function of pressure for ( open symbols ) and ( filled symbols , triangles : obtained from susceptibility data ; filled squares : obtained from raman data ) .
the solid lines give the predictions for the zero pressure derivatives . | pressure and temperature dependent susceptibility and raman scattering experiments on single crystalline have shown an unusually strong increase of the spin - peierls phase transition temperature upon applying hydrostatic pressure .
the large positive pressure coefficient ( 7.5 k / gpa ) - almost twice as large as for the pure compound ( 4.5 k / gpa ) - is interpreted as arising due to an increasing magnetic frustration which decreases the spin - spin correlation length , and thereby weakens the influence of the non - magnetic zn - substitution . version 2(june 23 1997 ) printed on |
let @xmath2 be the complete @xmath3-hypergraph on @xmath4 . in other words ,
let @xmath4 be a set and @xmath5 , where @xmath6 represents the family of subsets of size @xmath3 of @xmath4 .
a _ colouring _ of @xmath7 is simply a function @xmath8 .
we think of each @xmath9 as a colour , so @xmath10 has colour @xmath11 if @xmath12 .
if @xmath13 $ ] , then we call it a _ @xmath14-colouring_. a subset @xmath15 of @xmath4 is _ monochromatic _ if @xmath16 for some @xmath17 $ ] , and it is _ rainbow _
if @xmath18 for every pair of distinct elements @xmath19 .
ramsey s theorem @xcite guarantees the existence of monochromatic subsets in sufficiently large structures .
there is a quote by motzkin @xcite that describes this behaviour : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ `` complete disorder is impossible . ''
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the precise statement of the theorem is as follows . for @xmath3 , @xmath14 and @xmath20 ,
there is a number @xmath21 such that the following holds : if @xmath22 then for every @xmath14-colouring @xmath23 of @xmath7 there is a colour @xmath11 and a monochromatic subset @xmath24 with @xmath25 and @xmath26 . over the years
ramsey s theorem has proven to be one of the most useful theorems in combinatorics .
it has countless applications in geometry , number theory , algebra , computer science and many other fields .
for more details about ramsey theory we refer the reader to @xcite .
anti - ramsey theorems appeared in 1973 @xcite .
in contrast to the classical ramsey theory which seeks monochromatic subsets , anti - ramsey theory seeks _ rainbow _ subsets .
the following quote from @xcite describes this behaviour : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ `` complete disorder is unavoidable as well .
jungi , j. neetil , r. radoii .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ of course , rainbow subsets of a given size may not exist for arbitrarily large hypergraphs so some extra conditions are needed .
there are several types of anti - ramsey theorems . for a survey
, we refer the reader to @xcite . in this note
we study an anti - ramsey theorem by lefmann , rdl and wysocka @xcite and generalize it to theorems [ thm : main ] and [ thm : infinite ] below .
theorem [ thm : main ] has interesting applications .
we use it to obtain polynomial bounds for some problems in geometry and number theory , improving results from @xcite and @xcite . in order to state our results , we need the following definition .
a _ @xmath0-sunflower _ of @xmath2 is a family @xmath27 consisting of edges ( called _ petals _ ) whose intersection has ( at least ) @xmath0 elements .
-sunflower with @xmath28 petals in an @xmath29-hypergraph . ]
we are interested in colourings of @xmath7 in which there are no large monochromatic @xmath0-sunflowers .
[ thm : main ] let @xmath4 be a set and @xmath30 and @xmath1 be natural numbers .
let @xmath23 be a colouring of @xmath7 such that any monochromatic @xmath0-sunflower has at most @xmath1 petals .
then there exists a constant @xmath31 depending only on @xmath3 such that if @xmath32 satisfies @xmath33 there is a rainbow @xmath24 with @xmath34 .
the history of this theorem can be traced back to babai @xcite who proved the case @xmath35 .
later alon , lefman and rdl @xcite extended babai s result for @xmath36 , @xmath37 and @xmath38 .
lastly lefmann , rdl and wysocka @xcite proved it for @xmath36 , @xmath39 and @xmath40 .
the proof of theorem [ thm : main ] is very similar to the proof of theorem 3 in @xcite
so we only include a sketch of it in section [ sec : proofs ] .
we also state a version of theorem [ thm : main ] for sets with infinite cardinality .
it should be noted that it is not a direct consequence of its finite version , which only implies the existence of arbitrarily large rainbow subsets .
the proof of this theorem can also be found in section [ sec : proofs ] .
[ thm : infinite ] let @xmath4 be an infinite set and let @xmath23 be a colouring of @xmath7 with no monochromatic @xmath0-sunflower with @xmath1 petals . then there is an infinite rainbow subset of @xmath4 .
note that the condition on @xmath1 can not be relaxed to only forbid finite monochromatic sunflowers .
it is easy to construct an example showing this : for any edge @xmath10 of @xmath41 , define @xmath42 .
there are some interesting applications of theorem [ thm : main ] .
these are stated below and their proofs are included in section [ sec : applications ] .
we begin with applications to geometry , specifically to erds - szekerez type problems @xcite . here
the objective is to find , within a large set of points in general position , a subset of a given size with specific properties .
usually the existence of such a subset can be shown using ramsey s theorem but the bounds on the size of the original set are usually extremely large .
the first corollary gives polynomial bounds to three of these problems .
[ coro : geometry ] let @xmath43 and @xmath32 be a positive integers .
there exists a constant @xmath44 depending only on @xmath43 such that if @xmath45 is a set of points in @xmath46 for which @xmath47 then the following hold : 1 .
if no @xmath48 points of @xmath45 lie on a @xmath49-sphere , then there exists @xmath50 with @xmath32 points such that the circumradii of all the @xmath43-simplices defined by elements of @xmath51 are distinct .
if no @xmath52 points of @xmath45 lie on a hyperplane then there exists @xmath50 with @xmath32 points such that the volumes of all the @xmath43-simplices defined by elements of @xmath51 are distinct .
if no @xmath52 points of @xmath45 lie on a hyperplane then there exists @xmath50 with @xmath32 points such that the similarity types of all the @xmath43-simplices defined by elements of @xmath51 are distinct .
part ( 1 ) of this theorem improves the main result in @xcite , in which the bound @xmath53 was given for the case @xmath54 . part ( 2 ) , with @xmath54 , improves a bound given in @xcite that uses ramsey s theorem taking it down from @xmath55 to @xmath56 .
furthermore , this version generalises to any number of dimensions . as a second application
we show a result regarding polynomials .
[ coro : polynomial ] let @xmath57 be a field and @xmath58 $ ] be a symmetric polynomial of degree @xmath59 .
there exists a constant @xmath44 depending only on @xmath43 such that if @xmath60 and @xmath61 then we can choose a subset @xmath50 of size @xmath32 such that all the evaluations of @xmath62 in @xmath63 distinct elements of @xmath51 are distinct .
if the polynomial @xmath62 has degree @xmath64 then @xmath62 is simply the sum of @xmath63 variables .
an analogous theorem for sums of @xmath3 distinct elements in abelian groups also holds and has a similar proof , although this was already a consequence of the theorem in @xcite .
finally we consider @xmath65-sequences , also called sidon sequences ( see @xcite ) .
these are increasing sequences ( possibly finite ) of positive integers @xmath66 such that the sums @xmath67 are distinct for every @xmath68 .
an equivalent definition is that the differences @xmath69 are distinct for every @xmath68 .
[ coro : sidon ] there is a constant @xmath70 such that any set @xmath45 of positive integers with @xmath71 contains a @xmath65-sequence of length @xmath32 . for comparison
, it is known that for every @xmath72 there is a large enough @xmath32 such that the set @xmath73 contains a @xmath65-sequence of length @xmath32 .
as stated before , the proof of theorem [ thm : main ] is very similar to the proof theorem 3 in @xcite and only needs minor adjustments .
therefore , we only include a sketch of the main idea .
the proof uses a theorem by ajtai et al .
@xcite that gives a lower bound on the independence number of an _ uncrowded _ hypergraph .
an uncrowded hypergraph is one which has no cycles of length @xmath63 , @xmath74 or @xmath75 .
let @xmath76 be a colouring with no monochromatic @xmath0-sunflowers with more than @xmath1 petals .
we wish to find a large rainbow subset of @xmath4 . for convenience , denote @xmath77 .
we construct a @xmath78-hypergraph @xmath79 as follows : the vertices are @xmath4 and @xmath80 is an edge if and only if it contains two different sets @xmath81 with @xmath82 .
then an independent set on @xmath79 corresponds to a rainbow set of @xmath7 .
there is a simple probabilistic argument to get rid of all the edges of @xmath79 while keeping a good portion of the vertices , but this method yields the desired bound without the @xmath83 factor .
we chose to illustrate it here , as the proof of the tighter bound involves applying the same method to instead get rid of all 2,3,4-cycles of @xmath79 first while keeping a larger amount of vertices , and then using the lower bound on the independence number of an uncrowded hypergraph by ajtai et al .
the sunflower condition implies that the number of edges of @xmath79 can be bounded as follows : @xmath84 but these edges have size @xmath78 , which is larger than @xmath85 , since @xmath86 .
we use the probabilistic method : choose @xmath87 and then choose each vertex independently at random with probability @xmath62 ( erase with probability @xmath88 ) .
since @xmath89 , if we choose @xmath62 adequately we can `` destroy '' all but a few edges .
let @xmath90 denote the random variables which count the number of vertices and edges respectively .
by linearity of the expected value @xmath91 , we have : @xmath92 after this process we would like to obtain a hypergraph in which the number of edges is in @xmath93 , so we can delete the remaining edges `` by hand '' ( by , say , deleting one vertex from each ) .
this can be assured if the expected value of @xmath94 is in @xmath93 .
equality between @xmath95 and @xmath96 is achieved when @xmath97 , so that @xmath98 . choosing a slightly smaller @xmath62 achieves the desired result . with some technical details
omitted , we can get an independent set of @xmath79 of size in @xmath99 , which is a rainbow of the original .
the gritty details for the improved bound can be found in @xcite , and their proof works almost word for word for this generalisation , the only difference being that some expressions have to be multiplied by the constant @xmath1 , which does not affect the asymptotic behaviour .
now we prove the generalisation of theorem [ thm : main ] to infinite hypergraphs .
this theorem is implied by a theorem of erds and rado @xcite .
[ thm : er ] let @xmath100 be any colouring of the @xmath3-sets of @xmath101 .
then there exists an infinite subset @xmath102 and a set of indices @xmath103 with @xmath104 such that two ordered subsets @xmath105 have the same colour if and only if @xmath106 for every @xmath107 $ ] .
for example , when @xmath108 , we have a monochromatic structure because the colour of a @xmath3-set does not depend on any of the elements , and when @xmath109 $ ] we have a rainbow structure because the colour of a @xmath3-set depends on every element .
now we are ready to prove theorem [ thm : infinite ] . without loss of generality
we may assume @xmath110 .
suppose @xmath111 has the property that any monochromatic @xmath112-sunflower has at most @xmath1 petals .
let @xmath113 and @xmath114 $ ] be sets as in theorem [ thm : er ] , we wish to prove that @xmath109 $ ] .
for sake of contradiction , assume that some number @xmath115 $ ] is not in @xmath116 .
fix numbers @xmath117 in @xmath45 such that there are distinct @xmath118 strictly between @xmath119 and @xmath120 .
then the @xmath3-sets @xmath121 for @xmath122 form a monochromatic @xmath123sunflower with @xmath124 petals .
theorem [ thm : infinite ] makes it clear that the corollaries above have corresponding infinite versions .
in this section we will only use the case @xmath110 of theorem [ thm : main ] , so the condition on the size of @xmath4 becomes @xmath125 the proofs are straightforward , all that needs to be done is define a colouring of the @xmath3-tuples of a set and show that the monochromatic sunflower condition is satisfied for some @xmath1 . then theorem [ thm : main ] gives us the desired subset . in part
( 2 ) , we colour a @xmath127-tuple by its volume .
given @xmath43 points , the set of points that complete this @xmath43-tuple into a simplex of a prescribed volume is the union of two hyperplanes , since there are at most @xmath43 points on each , we set @xmath129 . for the last part , we colour a @xmath127-tuple by its similarity type . fix @xmath130 and let @xmath131 and @xmath132 be sets of @xmath52 points that determine a certain similarity type but with different orientations . if @xmath133 makes @xmath134 have the same similarity type as @xmath131 , then it determines a unique injective function @xmath135 or a unique injective function @xmath136 . since there are @xmath137 of these functions , we may take @xmath138 . since @xmath62 is symmetric we can colour a pair of distinct elements @xmath139 by the value of @xmath140 .
now write @xmath141 .
let @xmath142 be the smallest index for which @xmath143 , @xmath144 be the zero - set of @xmath145 , and @xmath146 .
note that , because the degree of @xmath62 is @xmath43 , @xmath144 has at most @xmath43 elements .
for every fixed @xmath147 , @xmath148 is a polynomial on @xmath149 with degree between @xmath64 and @xmath43 , therefore the equation @xmath150 has at most @xmath43 solutions .
thus , we can set @xmath151 . | a @xmath0-sunflower in a hypergraph is a family of edges with @xmath0 vertices in common .
we show that if we colour the edges of a complete hypergraph in such a way that any monochromatic @xmath0-sunflower has at most @xmath1 petals , then it contains a large rainbow complete subhypergraph .
this extends a theorem by lefmann , rdl and wysocka , but this version can be applied to problems in geometry and algebra .
we also give an infinite version of the theorem . |
the extreme carbon star irc+10216 is a classic example of a red giant caught in the act of evolving into a planetary nebula .
its relative proximity , high infrared luminosity , and abundance of molecules found in its dense outflow has resulted in a barrage of observations by astronomers , working across the spectrum , but particularly in the infrared and millimeter / sub - millimeter . despite all this attention , a good model of what is happening in the innermost regions where
the stellar outflow is born and accelerated is still sorely lacking .
numerous studies of molecular lines in the outer envelope ( e.g. @xcite ) have revealed a spherically expanding outflow , a finding which was beautifully confirmed with deep @xmath3 and @xmath4 band images of the dust shell in ambient scattered galactic light ( @xcite ) .
however this spherical symmetry , a characteristic of most red giant winds , will likely be broken as the irc+10216 evolves into a planetary nebula , most of which are elongated or bipolar ( e.g. @xcite ) .
the pronounced asymmetry in the innermost regions of the envelope of irc+10216 reported by numerous high - resolution imaging experiments ( most recently @xcite ; @xcite ; @xcite ; @xcite ) and also polarization studies ( @xcite ; @xcite ) suggests that the onset of this aspherical flow has already begun , probably within the last few hundred years . with a privileged vantage onto such a brief yet important period in the evolution of a low to intermediate mass ( initial mass @xmath55m@xmath6 @xcite ) star ,
high resolution observations are crucial in distinguishing between the many competing models for the physical mechanisms underlying the onset of asymmetry in the birth of a planetary nebula . in this paper
, we present a 7-epoch diffraction - limited imaging study of the inner dust shell of irc+10216 in the near - infrared k - band .
although some interpretation of the morphology of the images is given , full radiative transfer modelling results are beyond the scope of this report , and will be presented in a second paper .
instead , we emphasize here the detection and measurement of the motion of features presumably embedded in the outflow .
although proper motions have been reported for near - infrared images of dusty wolf - rayet shells ( @xcite ; @xcite ) , the two order - of - magnitude slower winds around asymptotic giant branch ( agb ) stars result in the requirement of longer time bases , and extremely high fidelity mapping schemes .
diffraction - limited images of irc+10216 were obtained at seven separate epochs with the keck i telescope , with dates and other observing details given in table 1 .
observations used the technique of aperture masking interferometry , by which starlight from the primary mirror is selectively blocked , with only a few regions of the pupil allowed to contribute to the final image . for observations of bright compact objects , of
which irc+10216 is a prime example , the methods of sparse - pupil interferometry have been shown to be fully competitive with , or superior to , other techniques such as speckle interferometry .
statistical methods based on the maximum - entropy technique ( @xcite ) have been used to recover maps from the complex visibility data ; however alternate methods such as the clean algorithm ( @xcite ) produced similar results .
data reduction and analysis procedures were also tested by observing a number of test objects , such as known binary stars , on each night .
a detailed description of the keck aperture masking experiment covering the observational techniques , data reduction , and image reconstruction can be found in tuthill et al .
( 2000 ) .
cllll 1 & 1997 jan 29 & golay15 & kcont & 0.72 1 & 1997 jan 29 & annulus & kcont & 0.72 2 & 1997 dec 16 & annulus & kcont & 1.23 2 & 1997 dec 16 & golay21 & kcont , ch4 & 1.23 2 & 1997 dec 18 & golay21 & kcont & 1.23 3 & 1998 apr 14 & golay21 & ch4 & 1.41 3 & 1998 apr 15 & annulus & ch4 & 1.41 4 & 1998 jun 04 & golay21 & ch4 & 1.49 5 & 1999 jan 05 & annulus & ch4 & 1.83 5 & 1999 jan 05 & golay21 & ch4 & 1.83 6 & 1999 feb 04 & golay21 & ch4 & 1.88 6 & 1999 feb 04 & annulus & ch4 & 1.88 7 & 1999 apr 25 & annulus & k , ch4 & 2.00 7 & 1999 apr 25 & annulus & ch4 & 2.00 at each epoch in table 1 , a number of observations of irc+10216 were made , often with quite distinct experimental setups .
three different aperture masks were used over the course of the project , with the non - redundant golay - type masks passing only a few percent , and the partially - redundant annulus mask passing around ten percent of the unobstructed pupil .
three different filters were also employed at various epochs , with the bandpass characteristics given in table 1 .
this diversity of observing parameters , while in part representing experimental evolution , allowed us to tailor the observations to conditions and specific requirements on a given night .
however , no systematic differences were found when comparing maps from annulus and golay data , and more significantly , from any of the different filters used ( filters differ in bandwidth , but have similar center frequencies ; c.f .
table 1 ) .
thus for the remainder of this work , all maps taken at one observing epoch are treated as measuring the same quantities .
two limitations of the interferometrically reconstructed maps should be mentioned .
firstly , the absolute photometry , or surface brightness scale in the maps , is difficult to calibrate with any great accuracy . for this reason ,
images shown have fluxes scaled relative to the peak intensity in each map , and discussion will refer to these relative fluxes .
secondly , the closure phase method does not deliver absolute positional information , and the center of each map has been chosen to be the location of the brightest pixel .
figure [ fig:6maps ] shows maps of irc+10216 from data spanning a period from 1997 january to 1999 april . in this figure , maps shown are the noise - weighted average over all image reconstructions from a given epoch ( or , in the case of jan / feb 99 , pair of epochs ) .
simple inspection shows that the basic morphology of the inner nebula surrounding irc+10216 has been very well established in the k band . from the most prominent structures down to
relatively minor features at only a few percent of the peak surface brightness , a high degree of consistency is found in the sequence of images .
furthermore , the early epoch images shown in figure [ fig:6maps ] are also in excellent agreement with recently published near - infrared images of weigelt et al .
( 1998 ) and haniff & buscher ( 1998 ) ( we refer collectively to these two papers as `` whb '' hereafter ) . in order to proceed with a quantitative analysis of the maps , a simple descriptive model has been used to identify and label features .
whb have labelled compact features a through d , with weigelt et al .
( 1998 ) adding e and f which seem much less distinct . with the considerably higher resolution available from the keck , these compact knots have in most cases been resolved into more complicated structures , and therefore a different approach is taken in describing the images , which we relate to the earlier schemes . a skeleton diagram of our model , based on analysis of images from figure [ fig:6maps ] , is given in figure [ fig : cartoon ] .
the location of the brightest feature in all maps , the relatively compact core , appears as a + surrounded by rings showing its approximate extent ( feature a of whb ) .
this core appears somewhat offset to the north - west from the center of an elongated , roughly elliptical region ( features e & f of weigelt et al .
( 1998 ) ) .
we refer to the core and these immediate surroundings as the southern component in figure [ fig : cartoon ] .
the next most prominent structure in the maps is the linear extension to the north and north - east , which we have split into two ; the shorter north arm ( whb feature c ) and the more prominent , elongated , north - east arm ( whb feature b ) .
the important north - east arm we have modelled as a linear ridge of emission . displaced to the east between these other features is a dimmer structure , consisting of multiple peaks we have labelled the eastern complex ( whb feature d ) . although signal - to - noise was not uniformly high enough to be assured of high - fidelity reconstructions of the eastern complex , for the purposes of characterizing the structure , we have labelled the most southerly portion cloud ec1 ( usually a little brighter and appearing like a ring or three peaks in a triangle ) , while in the more northern section we have been content to simply tag the location of the two most prominent peaks .
having established in the previous section a framework for describing the appearance of the maps of irc+10216 , we now proceed to examine the data for changes over the seven epochs .
the easiest thing to look for is a change in the relative positions of the components .
this was done with the use of a computer program which found the best - fit location of model components describing four features : the north - east arm , ec1 , and two minor peaks in the eastern complex .
as our maps have no associated astrometry , the registration between separate images is unknown and we have used the bright compact core as the fixed point against which to measure any motions . note that fits were not made directly to the averaged images presented in figure [ fig:6maps ] , but to the full 7-epoch dataset with multiple separate images at each epoch ( comprising a total of 25 maps ) .
this allowed errors to be determined on the locations of the features from the apparent spread of values over different maps .
all four features tracked were found to exhibit widening separations from the core as a function of time .
this can be verified by visual inspection of the images of figure [ fig:6maps ] where the northward motion of the north - east arm is readily apparent without any assistance from computer model - fits .
motions of the north - east arm and elements in the eastern complex including ec1 are shown in figure [ fig : cartoon ] . to avoid the confusion of seven sets of features on one plot , the motion is shown averaged into four time - intervals ( labelled t1 t4 ) , as described in the caption .
in addition to the motions of certain features , other changes are apparent with time in figure [ fig:6maps ] , at a level of significance well above the level of noise in the maps .
two of these in particular are worth highlighting .
the north - east arm which in 1997 january ends in a bright , fairly compact knot , evolves through a stage where it is of fairly uniform brightness along its length ( 1998 ) , and in the final measurements ( 1999 ) the extreme end of the arm has dimmed considerably .
the second interesting change concerns the central core , which in 1997 january appears fairly circular and compact , but by 1999 april clearly exhibits an extension to the south - east . further discussion of these points is given in section 3 below .
having established the presence of motion between the components , it is possible to characterize the apparent outflow of material by watching the features which are presumably embedded within it .
we restrict our attention to the north - east arm and ec1 , as the more minor features are too near to the map noise level to make useful tag points for the flow .
the first important question to be addressed is the assumption as to the origin of the flow .
whb present arguments that their component a , the compact core , be identified as the star itself .
certainly , features identified to be moving do appear to be moving away from a point to the south , and in our subsequent analysis we have assumed radial divergence from the compact core . however , there is no guarantee that any of the map features must be the stellar disk , which may lie behind some optically thick region , as is discussed further below .
it was fairly straightforward to project the motion of ec1 onto a vector beginning at the origin ( the brightest pixel in the core ; see figure [ fig : cartoon ] ) .
displacement along this vector with time then gives a velocity .
motion of the north - east arm , however , is not so easy to quantify .
the minimum possible apparent velocity consistent with the time - sequence of images would have material in the arm moving to the north - west , perpendicular to its length . while giving a lower bound on the possible velocity ,
this motion is not consistent with spherical outflow from our defined origin , and furthermore does not really describe the data as the arm would also need to grow in length . instead , we have labelled points ne1 & ne2 near the beginning and end of the arm respectively , and we measure the _ radial _ motion of these points from the origin .
since the arm has been fit as a linear ridge , two points are sufficient to completely determine its motion .
thus our three trace points in the flow are ec1 and the two points ne1 & ne2 on the north - east arm , all of which are assumed to be radially diverging from the core . plotted in figure [ fig : outflow ]
are the displacements from the origin of each trace point over the seven epochs . for the well - determined points
ne1 & ne2 lying on the high signal - to - noise north - east arm , the points can be seen to describe a very well defined linear relationship with time , which is echoed , with more scatter , by the ec1 data .
velocities may be computed from the slope of a least - squares fit to the data , with resulting apparent outflow rates given in figure [ fig : outflow ] .
the most direct way to observe an accelerating body is to follow its change of speed with time .
unfortunately , the error bars are too great , and the time - span of observation too short to make this strategy worthwhile for data in figure [ fig : outflow ] .
a uniform radial acceleration of 3.4mas.yr@xmath7 ( see below ) has been over - plotted ( dotted line ) illustrating the difficulty of measuring such a small curvature of the velocity law .
however , some information on the flow dynamics can be inferred from examination of the velocity field derived from the maps .
specifically , the greater the distance from the origin , the higher the outflow speed .
three possible explanations for this finding are described below .
the acceleration of material over the region sampled by the measurements may have been detected .
this acceleration can be easily visualized from the motion of the north - east arm : the ridge - lines of best fit are not parallel causing the arm to tilt northward with time . in order to do this , the far end of the arm ( ne2 )
must be moving faster than the near end ( ne1 ) .
the three flow - velocity data points from figure [ fig : outflow ] are consistent with a uniform radial acceleration of @xmath8mas.yr@xmath7 starting at a radius of @xmath9mas . as the principle acceleration mechanism is expected to be radiation pressure from the central star , a uniform acceleration law would not be expected
. however , in a region as clearly anisotropic as the inner dust shell of irc+10216 , flows may be complex and more realistic models will require greater efforts both in modelling and in recovery of longer time - baseline data .
a second possible explanation for the observed velocity field arises from the ambiguities associated with the projection of three - dimensional structure onto the plane of the sky .
such projected motions and velocities may give a misleading view of the true flow structure . to take an extreme example
, the north - east arm may , in reality , be a fragment of a circular arc , all of which is at a uniform distance from the star and moving at a uniform velocity .
however , when viewed from a relatively acute projected angle , apparent differences in separation and velocity will be recorded . for a roughly spherical distribution of clumps at a uniform flow speed ,
projection effects alone would result in slower apparent speeds being observed closer to the star , mimicking an acceleration .
the uniform ` acceleration ' law given above is appropriate for such a scenario , where the velocity will be proportional to the displacement from the origin .
finally , a third possibility which presents itself is that faster material is found further out ( e.g velocity law @xmath10 ) , but without acceleration taking place over the sampled region .
the most obvious origin for such a flow would be a stellar eruption ejecting material with a range of velocities at some point in the past , allowing faster clumps to move further from the star . under this scenario , back - projection of the flow through time , based on the ne1 & ne2 velocities
, implies that the north - east arm was created in an event around jd 2449000 ( about 1993 january ) originating from a point some 90mas from the core .
ideally , the assumption of a single expulsion event could be tested with three or more clumps by seeing if they appear to be diverging from a single point in space and time .
unfortunately , the errors on the ec1 flow are too large to offer a meaningful constraint , although a common origin for ne1 , ne2 & ec1 does fall within the errors . in summary , although acceleration stands as a strong candidate , the ambiguities of interpretation make it impossible to claim a clear detection .
indeed , some combination of all three effects discussed above may be needed to account for the true angular velocity field .
this highlights the need for high quality imaging over more extended periods to follow clumps from birth to dispersal in the extended shell . despite the uncertainties , a plane - of - the - sky assumption represents a simplest case and
should give , at the least , valid lower bounds to the velocity determinations discussed here .
taking the southern component and the north - east arm as the dominant bright structures in our images , the bipolarity at position angle @xmath2 reported throughout the literature is confirmed here .
however , at very high resolution , complex and clumpy structures are revealed which do not conform immediately to any simple axially symmetric models .
perhaps one of the most striking features of the maps of figure [ fig:6maps ] is the dark lane ( see figure [ fig : cartoon ] ) , around which all the major structures are distributed .
at all epochs , the flux level in this hole , in close proximity to all the brightest knots of emission , is consistent with a surface brightness of @xmath11% of the peak : similar to the noise level in the maps . that such a dark region could exist in the heart of the nebula argues for the presence of considerable material in the line - of - sight , leading to high levels of obscuration .
if this is the case , the interpretation of the maps becomes more difficult .
a bright knot might be either a hot clump of dust , or a `` window '' in the dust allowing a glimpse of the hot inner regions .
the modelling of the dust shell is a current work in progress , with the most promising interpretation being that the dark central band is a dusty torus or equatorial density enhancement , which we can see tilted towards us to the south revealing the inner hot regions ( southern component ) and possibly the star ( core ) .
clumpy features embedded in the outflow , such as the north - east arm and eastern complex , might have their origins in enhanced dust formation occurring above slowly - evolving massive convective features ( @xcite ) or magnetic spots ( @xcite ) on the stellar surface .
interestingly , no changes observed over the seven epochs , which cover more than one pulsational cycle ( @xmath0638days ) , gave clear evidence for new dust nucleation ( discounting the elongation of the core , discussed below ) .
this is in accord with model dust shells ( e.g. @xcite ) which show new layers of dust forming on timescales longer than one pulsation period .
until further modelling can be completed , we confine further discussion here to the motion of material in the outflow . for features such as the north - east arm , it seems most likely that motions detected are simply the displacement of emitting material , and not some more exotic scenario such as the motion of a viewing hole in the dust , or a warm spot caused by a ` searchlight beam ' where the star shines through a moving window in the dust .
the north - east arm does exhibit some common - sense characteristics reinforcing this view , such as the flux from the bright knot at its extreme end , initially bright , fading as it moves outwards and presumably further from the central star ( see section 2.3 ) .
perhaps the greatest uncertainties affecting the interpretation are the questions as to the origin and direction of the flow , and the three - dimensional structure of the nebula . although figure [ fig : cartoon ] does appear to show that features are moving away from a point in the general location of the southern component , it was not possible to pin this down precisely .
worse , it was not possible to tell if the southern component and core were moving also : it may be the case that all features are diverging from the star which lies obscured behind the dark central band .
a number of previous authors ( whb , @xcite ) have claimed that the central star is visible in near - infrared .
the compact feature we have labelled the core has an approximate angular size of @xmath12mas close to the expected angular diameter of the stellar photosphere ( @xcite ; @xcite ) .
however the progressive elongation of the core through time ( see section 2.3 ; figure [ fig:6maps ] ) is not easy to reconcile with this view .
it is possible that the south - east extension we see is simply the newest condensation of dust moving from the star , however for this to be the case the dust must be forming extremely close ( @xmath13 ) to the photosphere .
the inner radius of the dust shell has been recently estimated at @xmath14 ( @xcite ) and @xmath15 ( @xcite ) .
it may also be possible that a binary companion is playing some role in this inner distortion , although it is highly unlikely that direct light from a main - sequence star could have been seen . in this work
, the assumption has been made that the core does mark a fixed location associated with the star , with extra structure in the southern component arising from emission and/or partial obscuration from the inner boundary of the dust shell .
alternate scenarios , in which the core may also be moving , lead to modification of the derived velocities by up to a factor of 2 . the gas outflow velocity at large distances from the star is well established from co line profile studies to be 14.5km.sec@xmath16 , however as calculated by groenewegen ( 1997 ) , the dust will drift through the gas at 3km.sec@xmath16 resulting in a dust outflow speed of 17.5km.sec@xmath16 .
combining this with our maximum angular velocity of 25.5mas.yr@xmath16 from figure [ fig : outflow ] yields a distance estimate of 145pc to irc+10216 .
this is in accord with modern estimates lying in the range 110 170pc ( @xcite ; @xcite ; @xcite ) . however , there are too many uncertainties involved to place high confidence in this result .
in addition to the geometric ambiguities already mentioned , it is unclear if the dust clump followed here ( ne2 ) has finished accelerating and is at its terminal velocity .
furthermore , the value of v@xmath17 measured in the spherical molecular shell may not be a good measure of the inner dust motions .
it is difficult to imagine that any dramatic change from spherical outflow to an equatorial disk and bipolar lobes in irc+10216 did not also entail changes in the velocity structure . taking a distance of 145pc ,
proper motions of 11.5 , 17.8 & 25.5mas.yr@xmath16 ( figure [ fig : outflow ] ) imply outflow velocities of 7.9 , 12.2 & 17.5km.sec@xmath16 for points ne1 , ec1 , and ne2 respectively .
as computed above , the velocity structure is consistent with an apparent uniform acceleration of @xmath8mas.yr@xmath7 from rest at @xmath9mas .
however , the projection of a three - dimensional motion onto the plane of the sky , and unknowns associated with the initial conditions of each clump may also account for the range of observed flow speeds
. models of radiatively - driven dust acceleration predict a more complicated velocity law dependent upon many properties of star and the outflowing gas and dust ( @xcite ; @xcite ) .
however , detailed comparison with such predictions is premature until the velocity of a single feature can be shown to be changing over time .
diffraction - limited images recovered using interferometric techniques from a multi - epoch study spanning more than two years at the keck i telescope are presented . taken in the near - infrared k band
, the maps have revealed an asymmetric and clumpy structure at angular resolutions exceeding the expected diameter of the stellar photosphere ( @xmath0 50mas ) .
the most likely morphology for the circumstellar environment is an optically thick circumstellar disk or torus , possibly tilted towards the line of sight in the south revealing the hot inner cavity and emission from the stellar photosphere .
the angular separations of clumps of material thought to be in the northern bipolar lobe have been followed over time , revealing increasing separation from the compact core to the south .
outflow velocities derived from this motion are consistent with estimates of the radial outflow velocity ( from co measurements ) and the expected distance .
clumps at greater distances from the core were found to show increasing velocities , which may be taken as evidence for acceleration in the inner regions ; the effects of geometrical projection ; or the result of a past event which ejected material with a range of velocities .
in addition to the changing separations of components , the appearance of the inner nebula was found to be evolving in other ways , however none were interpreted as evidence for new dust condensation over the pulsation cycle .
further modelling of this system is currently underway , and will be presented in a subsequent paper .
we would like to thank the referee , matt bobrowsky , for helpful suggestions and devinder sivia for the maximum - entropy mapping program `` vlbmem '' .
data presented herein were obtained at the w.m .
keck observatory , which is operated as a scientific partnership among the california institute of technology , the university of california and the national aeronautics and space administration .
the observatory was made possible by the generous financial support of the w.m .
keck foundation .
this work is a part of a long - standing interferometry program at u.c .
berkeley , supported by the national science foundation ( grant ast-9315485 and ast-9731625 ) by the office of naval research ( ocnr n00014 - 89-j-1583 ) , and the france - berkeley fund . | the results of a 7-epoch interferometric imaging study , at wavelengths in the near - infrared k - band , of the carbon star irc+10216 are presented .
the use of non- and partially - redundant aperture masking techniques on the 10-m keck - i telescope has allowed us to produce images of the innermost regions of the circumstellar dust envelope with unprecedented detail . with roughly twice the resolving power of previous work ( @xcite ;
@xcite ; @xcite ) , the complex asymmetric structures reported within the central have been imaged at the size scale of the stellar disk itself ( @xmath050mas ) . a prominent dark lane at a position angle of approximately @xmath1
is suggested to be an optically thick disk or torus of dust which could help to explain irc+10216 s well - known bipolarity at a position angle of @xmath2 .
observations spanning more than a pulsational cycle ( @xmath0638days ) have revealed significant temporal evolution of the nebula , including the outward motion of bright knots and clumps .
registering these displacements against the compact bright core , which we tentatively identify as marking the location of the star , has allowed us to determine the apparent angular velocity at a number of points .
the magnitudes of the proper motions were found to be in agreement with current estimates of the stellar distance and radial velocity .
higher outflow speeds were found for features with greater separation from the core .
this is consistent with acceleration taking place over the region sampled by the measurements , however alternate interpretations are also presented .
although a number of changes of morphology were found , none were clearly interpreted as the condensation of new dust over the pulsation cycle .
unfortunately , ambiguities associated with the true three - dimensional nature of the nebula weaken a number of our quantitative and qualitative conclusions . |
stars with initial masses between 0.8 and 8 m@xmath2 pass through the asymptotic giant branch ( agb ) phase of evolution , and undergo significant mass loss @xcite . evolving from the tip of the agb to the planetary nebula stage , stars move through the post - agb phase ( typically 10@xmath3
10@xmath4 years ) , during which the circumstellar dust shell moves away from the star and decreases in temperature , from @xmath5 400 to 100 k @xcite .
the properties of post - agb stars have been reviewed by , for example , @xcite and @xcite .
this evolutionary phase is important as it provides an insight into nucleosynthesis and mixing processes during the later stages of stellar evolution .
most post - agb stars have been identified from their infrared excess ( e.g. , from _ iras _ observations ) , originating from the circumstellar dust shell .
this shell is more readily detected during the initial stages of post - agb evolution , when it is relatively hot and close to the star , and hence the majority of stars identified as being at a post - agb evolutionary stage have just left the agb and are of spectral type a , f or g @xcite .
however , there are many features of post - agb evolution which are not yet fully understood . for example , several stars exhibit abundance patterns similar to interstellar gas , with elements such as mg , si , al , ca , ti etc following the very low fe abundance , while c , n , o and s are near solar ( @xcite ; see , for example , ) .
@xcite discussed the conditions necessary for such a ` cleaned - up ' photosphere in post - agb stars , suggesting that the process is likely to occur during mass transfer between members of a binary system , or accretion triggered by a binary companion .
@xcite added that accretion of material from a circumstellar disc around a binary system was a favorable scenario . in this , grains form in the circumstellar shell produced during the agb phase , and the subsequent slow accretion of gas from the material results in the observed abundance patterns .
@xcite found that all the extremely iron deficient post - agb stars , known at that time , were binaries , supporting this hypothesis .
although the majority of post - agb candidates are cool objects , several b - type stars originally classified as population i stars are now believed to be at this evolutionary stage , on the basis of their derived atmospheric parameters and chemical compositions ( see , for example , @xcite ) .
these b - type stars should be evolving from the cooler post - agb stars , but significant differences are displayed .
the most obvious is the large carbon underabundance found for several hot objects , which may imply that they have left the agb prior to the onset of the third dredge - up ( for example , ) .
the abundance differences between hot and cool post - agb candidates indicate that further study is required to fully understand this evolutionary stage .
it is important to begin by studying hot post - agb objects of known initial metallicity , rather than field stars of unknown origin .
such a sample are the ultraviolet ( uv)-bright stars found in several globular clusters , whose post - agb nature has been established from their position in the cluster colour magnitude diagram @xcite .
the term ` uv - bright ' was first suggested by @xcite and denotes stars which have ultraviolet fluxes larger than those in either the red giants or the horizontal branch stars @xcite . for an excellent review of hot stars in globular clusters
see @xcite .
roa5701 , in the galactic globular cluster @xmath0 centauri , is a well studied uv - bright star due to its relative brightness ( ) .
the metal - poor @xmath0 cen is known to have ( at least ) three stellar populations , with metallicities : [ fe / h ] @xmath5 1.6 , corresponding to the metal poor population containing the bulk of the stars , [ fe / h ] @xmath5 1.2 for an intermediate metallicity population and [ fe / h ] @xmath5 0.5 for a metal - rich population comprising 5 per cent of the stars ( ) . in this paper , we analyse high - resolution optical spectra of roa5701 using non - local thermodynamic equilibrium ( non - lte ) model atmosphere techniques to estimate its atmospheric parameters and chemical composition , leading to a reconsideration of its evolutionary status .
the spectroscopic data presented here were obtained with the 3.9-m anglo australian telescope ( aat ) on the nights of 2000 july 15 and 16 .
the university college london chelle spectrograph ( ucls ) was used with the mitll 2a ccd ( 2048 @xmath6 4096 ) detector , the 31 groove @xmath7 grating and the 70 cm camera , giving a linear dispersion of 2 @xmath7 and a spectral resolution [ full width at half maximum ( fwhm ) ] of _ r _ @xmath5 40000 .
the read - out noise and gain were 1.8 @xmath8 and 0.37 @xmath9 respectively .
two overlapping wavelength regions were observed , resulting in a spectral coverage of 37609400 .
flat - field and bias frames were exposed at the beginning and end of the nights , while thorium - argon wavelength calibration spectra were interweaved with the stellar spectra .
reduction of the raw two dimensional ccd images were performed using standard procedures within the image reduction and analysis facility ( iraf ; @xcite ) .
procedures such as bias subtraction and flat fielding were achieved using the ccdred package @xcite .
cosmic - ray removal , sky subtraction , stellar extraction and wavelength calibration were performed using specred @xcite and doecslit @xcite .
after extraction , the spectra were co - added ( resulting in a signal - to - noise ratio of @xmath5 30 ) and inserted into the starlink package dipso @xcite for further analysis .
the spectral continuum was normalised , with low - order ( @xmath10 5 ) polynomials in regions of spectra free from absorption lines . following this , they were radial velocity shifted to the restframe , using unblended metal and hydrogen lines .
this resulted in the kinematical local standard of rest radial velocity @xmath11 = + 246 @xmath12 3 @xmath13 , using corrections generated by the starlink progam rv @xcite .
equivalent widths were measured for metal absorption lines by utilising the emission line fitting ( elf ) routines within dipso .
elf uses non - linear least squares gaussian fitting , with line centre , linewidth and line strength as variable parameters .
the typical fwhm for the metal absorption line spectrum was @xmath5 11 @xmath13 , indicating that a small amount of broadening is evident in the lines .
the error associated with the measurement of equivalent widths has been previously discussed @xcite , with a well - observed unblended feature having an accuracy of typically 10 per cent ( designated as quality a in table [ tab_all ] ) , a weak or blended feature accurate to 20 per cent ( designated b ) , and weaker features with accuracy less than 20 per cent ( designated c ) . for the balmer hydrogen lines ( i.e. , h@xmath14 , h@xmath15 , h@xmath16 , h@xmath17 ) and diffuse hei lines , equivalent widths were not measured , but theoretical profiles were fitted to the normalized observational data ( see , for example , fig . [ fig_h ] ) . at 4861 in the spectrum of roa5701 .
the theoretical profiles have been generated for surface gravities of 3.0 , 3.25 and 3.5 dex ( lower , middle and upper smooth curves , respectively ) , and imply a surface gravity of log _ g _ = 3.2 dex.,scaledwidth=50.0% ] .equivalent width measurements and corresponding abundances for the observed metal lines . [ cols="<,^,^,^,^,^,^,^,^",options="header " , ] studies of post - agb stars have shown that some of these objects exhibit abundance trends similar to that of interstellar gas , i.e. , with cno and s near solar and mg , si , ca , etc following the low abundances of fe @xcite .
examples of such objects include hd46703 @xcite and hd56126 @xcite .
grain formation in the circumstellar shells of stars during the agb evolution has been suggested to explain the origin of such a ` cleaned - up ' photosphere .
the premise is that during the agb phase , in the cool atmosphere of the star , a circumstellar shell is formed in which dust condenses .
dust particles are removed by radiation pressure , while the ` cleaned ' gas is re - accreted onto the stellar surface .
iron , and similar metals , have a high condensation temperature , so are more likely to condense onto dust grains .
elements such as c , n and o , with lower condensation temperatures , show no significant depletions and stay in the gas phase @xcite .
hence , the remaining re - accreted gas forms an iron - poor atmosphere .
@xcite suggest that roa5701 has undergone this gas dust separation as they found a very low fe abundance , with n and o relatively undepleted .
however , in this scenario one would expect fe to have a similar depletion to mg and si , as these would also be subject to grain formation @xcite . from our work
we find that mg and si are underabundant by [ mg / h ] = 1.1 @xmath12 0.2 dex and [ si / h]= 1.3 @xmath12 0.3 dex ( relative to the ism ; table [ tab_ay ] ) .
however @xcite derive a fe abundance of [ fe / h ] = 2.6 dex ( relative to the ism ) from _ hubble space telescope _
( _ hst _ ) spectra , which is more than 1.0 dex lower than the mg and si values we find .
also , we have derived an abundance for s of [ s / h ] = 1.4 @xmath12 0.2 dex ( relative to the ism ) , so the mg , si and s values agree with each other and are consistent with the average metallicity of the cluster , namely [ fe / h ] = 1.2 dex @xcite .
this result would remain unchanged if we had adopted abundance estimates relative to those from young b - type stars . if grain formation has occured , one would expect s to be relatively undepleted and therefore not have an adundance pattern in agreement with those of mg and si , thus implying that , for roa5701 , there is no evidence of a ` cleaned - up ' photosphere .
there are two possibilities to explain the abundance patterns seen .
either the iron abundance suggested by @xcite is incorrect , or it is indeed very low and an alternative mechanism is needed to explain the abundance results found .
one explanation for the low fe abundance is s - processing , with fe reduced via successive neutron capture of @xmath18fe .
this scenario has been applied to h - deficient very hot post - agb stars , e.g. , pg1159 type star k1 - 16 .
however s - process enriched objects are known to cover a range of metallicities between [ fe / h ] = 0.2 to 1.0 dex @xcite , much greater than the value @xcite find for roa5701 , i.e. , [ fe / h ] = 2.7 dex .
also , this mechanism is normally observed in post - agb stars which have c - rich circumstellar material , such as hd158616 @xcite . by contrast , we have found c to be deficient by [ c / h ] @xmath10 1.4 dex and so s - processing seems unlikely .
the low c abundance could be explained by hot bottom burning ( hbb ; @xcite ) , as suggested for v453oph @xcite . in this , temperatures become large enough for cno processing to begin at the base of the convective envelope , with @xmath19c converted to @xmath20n and moved to the surface .
@xcite suggests that hbb may have an important role at extremely low metallicities .
only population i stars of initial mass m @xmath21 5 m@xmath2 , and population ii stars of m @xmath21 3.5 m@xmath2 , experience hbb @xcite , and as the initial mass of roa5701 is unknown , no conclusions can be made regarding this process .
the fe abundance pattern observed in roa5701 does not appear to be plausible in terms of s - processing .
for the abundance patterns of the other elements , @xcite note that it is not necessary to invoke dredge - up during the agb phase to explain these , but a moderate third dredge - up can not be ruled out . the processes of dredge - up are well known ( see , for example , @xcite ) .
briefly , dredge - up occurs during the stars evolution as the products of hydrogen and helium burning are transported from the core to the surface , changing the surface abundances .
at the first dredge - up , the surface @xmath20n abundance is increased at the expense of @xmath19c , while the @xmath22o abundance sees little change . during the second dredge - up
, hydrogen products are again moved to the surface , with @xmath19c and @xmath22o almost completely converted to @xmath20n .
the third dredge - up sees the products of the helium burning cycle , following the triple@xmath14 process , brought to the surface . as a result , @xmath19c is enhanced relative to @xmath20n and @xmath22o , @xmath20n being formed during the hydrogen burning and @xmath22o formed through @xmath14capture by carbon atoms .
also s - process elements are dredged - up to the surface , and so would provide evidence for the third dredge - up @xcite .
objects which have undergone the third dredge - up would , therefore , have c - rich material , for example k648 in the globular cluster m15 @xcite .
hotter post - agb objects display a deficiency in c , unlike cool post - agb stars , which suggests that they must evolve off the agb before the third dredge - up begins , as the final convection would bring the products of helium burning to the surface @xcite .
carbon deficiencies have been observed in hotter objects with a lack of third dredge - up suggested , for example , zng1 in the globular cluster m10 @xcite , barnard 29 in m13 @xcite and hd341617 @xcite .
roa5701 follows the trends seen in typical hot , post - agb stars .
the c abundance is [ c / h ] @xmath10 1.4 @xmath12 0.2 dex , with o and n underabundant by [ o / h ] = 0.8 @xmath12 0.1 dex and [ n / h ] = 0.5 @xmath12 0.2 dex , relative to young b - type stars ( table [ tab_ay ] ) .
hence n , and o to a lesser extent , are enhanced relative to the other elements and the average metallicity of the cluster .
@xcite observe a similar enhancement of n and o relative to the other elements .
however their n abundance is not increased relative to o probably because they observed fewer nii lines . for c
, only an upper limit could be set for the abundance , which indeed is lower than the average metallicity of the cluster .
the c , n and o abundances measured here imply that roa5701 has undergone the first and second dredge - ups as the hydrogen burning products appear to have been brought to the surface .
therefore , the c underabundance suggests that , as with other post - agb stars , roa5701 has left the agb before the onset of the third dredge - up .
we have investigated whether roa5701 could be an agb manqu star originating from the extreme horizontal branch ( ehb ) , with diffusion still active and causing the low iron abundance observed by @xcite .
@xcite uses the terminology ehb to describe hb sequences of constant mass that do not reach the thermally pulsing stage on the agb . instead ,
following core helium exhaustion , they evolve into either post early agb stars , which leave the agb before thermal pulsing , or agb manqu stars , which never develop extensive outer convective zones and so do not reach the agb .
ehb stars have envelope masses that are too small to reach the end stages of normal agb evolution .
@xcite give the luminosity of roa5701 as log ( l / l@xmath2 ) = 3.2 , which combined with the evidence for the occurance of the first and second dredge - ups implies that roa5701 has evolved along the agb .
therefore , roa5701 does not appear to be an agb manqu star , so the low iron abundance can not be explained via this scenario .
however , it is reasonable that roa5701 could be a post early agb star as there is no evidence for the occurance of the third dredge - up , see for example phl1580 and phl174 @xcite , and id6 in ngc5986 @xcite .
assuming that the star has not undergone the third dredge - up or followed a gas dust separation , the origin of the low fe abundance suggested by @xcite remains unclear .
hence a more detailed investigation into this abundance is required .
as stated previously , it is possible that with higher signal - to - noise optical spectra feiii lines could be detected .
we plan to seek such observational data in the future and examine the _
hst_-uv spectra to improve upon the exisiting iron abundance estimate .
we are grateful to david pinfield , fergal crawford , and the staff of the aat for obtaining the observations .
we would like to thank ian hunter for his assistance with data reduction and analysis , and carrie trundle for her help with line identification .
we thank the anonymous referee for their helpful comments .
acknowledge financial support from the northern ireland department of education and learning ( del ) .
f.p.k . is grateful to awe aldermaston for the award of a william penney fellowship .
landsman w. , moehler s. , napiwotzki r. , heber u. , sweigart a. , catelan m. , stecher t. , 2000 , in leige , noels a. , magain p. , caro d. , jehin e. , parmentier g. , thoul a. a. , eds , 35th leige int .
colloq . , the galactic halo : from globular cluster to field stars .
et de geop . ,
belgium , p. 515 | an analysis of high - resolution anglo - australian telescope ( aat)/ university college london chelle spectrograph ( ucls ) optical spectra for the ultraviolet ( uv)-bright star roa5701 in the globular cluster @xmath0 cen ( ngc5139 ) is performed , using non - local thermodynamic equilibrium ( non - lte ) model atmospheres to estimate stellar atmospheric parameters and chemical composition .
abundances are derived for c , n , o , mg , si and s , and compared with those found previously by moehler et al . we find a general metal underabundance relative to young b - type stars , consistent with the average metallicity of the cluster .
our results indicate that roa5701 has not undergone a gas dust separation scenario as previously suggested .
however , its abundance pattern does imply that roa5701 has evolved off the agb prior to the onset of the third dredge - up .
[ firstpage ] stars : abundances stars : agb and post - agb stars : early - type stars : individual : cl@xmath1ngc5139wor1957 globular clusters : individual : ngc5139 |
a significant amount of gas processed by stars may be ejected into interstellar space via stellar winds and supernova explosions .
about 30% of the metals in enriched gas would condense to form dust grains @xcite .
the total amount of dust in the universe that is produced in stellar evolution in the entire cosmic time was estimated from integrated star formation rate by @xcite to be @xmath4 , in units of the present - day critical mass density .
this value is a few times higher than the amount of dust observed in galactic disks @xcite , which motivated us to explore the fate of the remaining amount .
evidence for the existence of dust beyond galactic disks has observationally been suggested by a few authors . from the study of a low - redshift foreground / background galaxy superposition
@xcite detected dust extinction up to about five times the optical extent of spiral galaxies . using deep _
herschel _ observations of m82 , showed that emission from cold dust can be traced up to 20 kpc from the centre of the galaxy .
recently , ( * ? ? ?
* hereafter msfr ) measured the cross - correlation between the colors of distant quasars and foreground galaxies as a function of the angular separation to galaxies , and concluded an excess reddening signal on scales ranging from 20 kpc to a few mpc , implying the existence of a appreciable amount of dust in intergalactic space .
using this observational result , @xcite showed that the summed contributions of dust in and outside galaxies appear to be in agreement with the total amount of dust ought to be produced in the universe .
this implies that dust destruction does not play a major role in the global dust distribution and that most of the intergalactic dust survives over the cosmic time . in this paper
we pursue another line of observations and show that a significant amount of dust resides in galactic halos and possibly beyond : we use absorbers to find dust contained therein .
is the most commonly detected absorption line from cool gas ( @xmath5 k ) at @xmath6 in the optical spectra of distant sources .
strong absorbers , conventionally defined with a rest equivalent width @xmath7 , are associated with a range of galaxies with @xmath8 and are found at impact parameters ranging up to 100 kpc @xcite . while the physical mechanisms for the origin of these gas clouds are yet to be understood , most absorbers seem to reside in galactic haloes .
the use of absorbers offers an attractive property : the knowledge of the absorber incidence @xmath9 allows us to infer the cosmic mass density of dust contained in these systems without further assumptions on their spatial distribution .
absorbers can hence be used to obtain a robust lower limit on the amount of baryons and dust residing outside galactic disks up to the halo radius , over the redshift range @xmath10 . in general , it is widely believed that absorbers do not host star forming regions and hence no source of dust grains .
the presence of dust in these clouds would serve as an indicator to distinguish whether absorbers predominantly are aggregates of pristine gas or are of secondary products from activities of nearby galaxies . characterizing and understanding its distribution is therefore an important task .
the use of absorbers also allows us to infer the neutral hydrogen abundance in clouds .
these pieces of information concerning the ingredients of the clouds would lead us to infer properties of absorbers , and then the nature and the formation of these systems . at the same time
this study enables us to explore the feature of the extinction curve to the short wavelength due to the high redshift nature of clouds , even to beyond the lyman limit , using optical and uv data currently available .
our basic data are taken from the sdss ( york et al .
2001 ) but are supplemented with uv photometry of the galaxy evolution explorer ( galex ; martin et al .
we use @xmath11 km s@xmath12mpc@xmath12 and @xmath13 in a flat universe .
metal - enriched gas causes extinction of light from uv to near infrared passing through intergalactic matter due to absorption and scattering by dust grains . in the far uv region
we also anticipate that ionisation and excitation of hydrogen atoms lead to an attenuation of light .
this is usually not addressed in the context of dust extinction observed in optical wavelengths . in a similar manner excitation of heavy elements may also contribute to the extinction opacity .
we write the optical depth for extinction as @xmath14 \label{eq : tau}\ ] ] where @xmath15 is the hydrogen cross section , @xmath16 is the dust extinction cross - section per hydrogen atom for a given population of dust grains denoted by @xmath17 , @xmath18 accounts for variations in the dust - to - gas ratio , @xmath19 is the metal excitation cross section and @xmath20 is the metallicity in the ratio of numbers of atoms . figure [ fig : cross - section ] indicates the cross - sections as a function of wavelength from the uv to the optical regime in ( a ) , and the representative transmission for a cloud with hydrogen column density @xmath21 in ( b ) and ( c ) .
dust models for the milky way and the smc are adopted from @xcitedraine / dust / dustmix.html ] .
these models , consisting of mixture of carbonaceous grains and astronomical silicate grains , reproduce the observed extinction curves from the ultraviolet to the near infrared in the milky way and magellanic clouds .
they are normalized such that @xmath22 and @xmath23 , respectively , where @xmath24 stands for extinction in the @xmath25 band .
the hydrogen cross - section is shown for gas at a temperature @xmath26k , but other choices do not modify the resonant feature qualitatively .
there are humps at @xmath27 m and @xmath28 m in the dust extinction curve caused by carbonaceous grains , though they are squeezed to a barely recognizable level in the top panel of the figure due to the broad logarithmic scale .
the transmission @xmath29 shown in the lower panels of figure [ fig : cross - section ] assumes the neutral hydrogen column density @xmath30 , which corresponds roughly to the median column density of absorbers with @xmath31 @xcite .
equation [ eq : tau ] shows that we can obtain , in principle , information on @xmath32 and @xmath33 from the brightness of a source located behind such a cloud .
the figure shows that atomic excitation not only at the lyman edge but also in the lyman region is a source of optical depth .
a cloud of this hydrogen column density is completely opaque at wavelengths shorter than the lyman limit , @xmath34 . at longer wavelengths the optical depth is @xmath35 , which is dominated by dust extinction .
a drop at 2175 is seen in the transmission through milky way type dust . and @xmath26k .
the blue curve shows a composite metal absorption spectrum for absorbers . _
bottom : _ transmission through milky way type and smc type gas with the corresponding dust - to - gas ratios . ] as a function of redshift , using the smc type dust model from @xcite .
breakdown into hydrogen , metal excitation and dust is shown in the three lower panels .
seven colour pass bands are shown with different colours indicated in the legend . ]
we show an example of extinction for observations conducted with broad band filters in figure [ fig : mag_shift ] .
the upper panel shows the attenuation of the flux induced by smc - type dusty gas of hydrogen column density @xmath36@xmath37 , as a function of redshift for the sdss ( @xmath38 ) and galex ( nuv , fuv ) passbands .
individual contributions are shown in the lower panels : ( b ) hydrogen , ( c ) metal lines , and ( d ) dust grains .
the seven passbands are denoted by different colours , as specified in the legend of the figure .
lyman - edge absorption dominates in the fuv , nuv and @xmath39 bands above redshifts 0.5 , 1 and 2.5 , respectively , as seen when one compares ( a ) and ( b ) .
we include atomic excitations from metal lines using a composite absorption spectrum for clouds .
the total contribution to the absorption opacity , which is dominated by , feii and civ lines , can be comparable to that of lyman-@xmath40 .
the 0.01 mag hump seen in the @xmath39 band at @xmath41 is due to lyman line absorption and the feature at @xmath42 is due to absorption by metals lines .
the lyman absorption is also seen in the fuv and nuv passbands , although the feature is somewhat weakened in nuv due to its larger width of the bandpass ( @xmath43 ) . at @xmath44
we see that lyman edge absorption becomes stronger in the nuv than the fuv band .
this causes a blueing effect in these bands .
similarly , this is expected for the nuv@xmath45 color at @xmath46 .
absorption due to metal excitation is also visible in the curves for @xmath47 passbands .
we see that metal excitation increases the extinction opacity from dust alone by as much as 5 times in the relevant wavelength range in all optical passbands . with the inclusion of metal excitation opacity
the extinction curve is not always monotonic , and the colour indices receive occasionally blueing for some wavelength ranges rather than reddening , with an amplitude of the order 0.01 magnitude . choosing milky way type dust in our demonstration would increase the contribution from dust opacity by the relative @xmath48 ratio , i.e. by roughly a factor 8 . here , our choice of the smc type dust is motivated from the fact that it fits the extinction spectrum of absorbers better , especially at short wavelengths @xmath49 , that is important to our consideration , and for the absence of the @xmath502175 feature @xcite .
when interpreting observations we leave , however , the dust - to - gas ratio as a free parameter , rather than fix it to the actual smc value .
this ratio is constrained from both shape of the extinction curve , for extinction is caused by both lyman edge absorption and dust scattering , and amount of dust and hi gas associated with absorbers .
previous analyses based on optical data have constrained the amount of dust in intervening absorbers by statistically measuring color changes they induce on background sources @xcite . here
we extend to the uv , which enables us to obtain information also on the amount of neutral hydrogen associated with absorbers .
the sdss ( york et al . 2001 ) provides us with the basic data set for the present analysis .
@xcite constructed a catalogue of absorbers using its edr data base @xcite , and @xcite extended it to the dr4 dataset @xcite .
@xcite analyzed about 45,000 quasar spectra and identified about 17,000 absorbers .
@xcite estimated that the redshift path covered by the survey drops to about 50% for absorbers with the rest - system equivalent width @xmath51 . while completeness is not an important issue in our analysis ,
we restrict our study to the sample of absorbers with @xmath52 .
we do not include absorption lines located close to the edges of the spectra and restrict the absorber redshift distribution to @xmath53 .
it is known from studies of absorber - galaxy pairs ( e.g. , steidel et al .
1994 ) that mgii absorber is unlikely to be a part of galactic disks , but are placed substantially away from galaxies . about 70% of the sdss quasars are observed by galex ( martin et al . 2005 ) .
budavari et al . (
2009 ) cross - matched the sdss - dr7 catalog with the galex - gr4/5 taking a matching radius of 4@xmath54 . using this cross - identification
, we can find 11,929 absorption systems satisfying our rest equivalent width and redshift selections and for which uv observations of the background quasars are available .
this constitutes the prime sample of our analysis .
, as a function of redshift for fuv@xmath55 , nuv@xmath55 , @xmath56 , @xmath57 , @xmath58 and @xmath59 colours in our quasar sample .
the solid lines are reddening expected for clouds with smc type dust having @xmath60 @xmath37 but scaled to the dust - to - hydrogen ratio of 1/100 consistent with our determinations given later .
dashed curves are expectation for @xmath61 @xmath37 clouds . ] for absorbers at @xmath62 and 2 ( solid circles ) .
reddening is evaluated at the median wavelengths of the 5 filters shorter than the @xmath63 band .
the dashed curve is the dust extinction curve of @xcite for smc , and dotted is absorption by hydrogen .
the two curves are added to give the total extinction curves shown by thick solid curves .
the curves assume @xmath60@xmath37 with the dust - to - gas ratio 1/100 .
the hydrogen absorption is evaluated for @xmath61@xmath37 . ] to detect dust in absorbers we measure reddening in broad band brightness correlated with the presence of absorption systems in quasar spectra . for each quasar with a detected absorber
we select four nearest neighbour quasars in redshift@xmath64-band magnitude space and take them to be reference quasars .
the high density of points in this space allows us to find reference quasars within a brightness difference typically smaller than 0.1 mag in the @xmath64-band .
we then compute a median colour excess and estimate its error by bootstrapping the absorber sample .
figure [ fig : color_redshift ] shows the reddening due to all absorbers with @xmath65 as a function of redshift . for six colours , fuv@xmath55 , nuv@xmath55 , @xmath56 , @xmath57 , @xmath58 and @xmath59 , the solid curves show the expected reddening including hydrogen absorption for log @xmath66 with smc - type dust but with a dust abundance lowered by a factor 6 , corresponding to a dust - to - gas mass ratio of about 1/108 our reddening estimation also includes absorption from metal lines .
the overall agreement is good : the deviations from the theory curve are of the order of a few percent except for some specific case , which we discuss below .
this indicates that the hydrogen column density and dust to gas ratio we used here are roughly correct without further adjustment .
we note that attenuation in @xmath58 is a very small quantity , with an amplitude smaller than @xmath67 mag .
interestingly , the broad - band photometric data indicates the presence of absorption lines , as shown by the bumps at @xmath68 and @xmath69 for the @xmath58 and @xmath59 colours . the small offsets ( of order 0.005 magnitude ) between observations and the modelled colours might be due to variations in the metal lines as a function of redshift .
we note that if we would take the mw type dust , instead of the smc type dust , we can not reproduce colour excess shown in figure 3 , notably in the @xmath56 for @xmath70 and @xmath57 for @xmath71 .
this is consistent with what has been found in york et al .
( 2006 ) , and compel us to adopt the smc type extinction for absorbers . in the fuv quasars behind a hydrogen column density with log @xmath72
are usually not detectable by galex .
the colour excess can only be measured for lower column density absorbers .
we also show in figure [ fig : color_redshift ] expected reddening for both absorbers with a column density somewhat above the lyman limit : log @xmath73 , as dashed lines .
the overall shape of the extinction shows that this selection effect is well reproduced .
figure [ fig : reddening_shape ] presents the extinction curve derived from our sample ( data points ) at two redshifts , compared with the one ( thick solid curve ) expected from a sum of the dust extinction curve ( dashed curve ) taken from @xcite and hydrogen lyman edge excitation ( dotted curve ) .
the curve assumes smc type dust for hydrogen column density @xmath74 @xmath37 with the dust - to - gas ratio 1/108 . for the hydrogen excitation
the hydrogen column density is assumed to be @xmath75 @xmath37 , consistent with figure [ fig : color_redshift ] , avoiding the transmission being completely opaque .
they are close to the case observed for our sample , as seen above .
the figure shows that the smc extinction curve gives observed extinction with absorbers correctly for @xmath76 m : extinction at shorter wavelengths is properly given by lyman edge absorption .
let us focus on dust extinction .
the above figure shows that over the redshift range @xmath77 extinction in the fuv , nuv and @xmath39 band is sensitive to hydrogen absorption . in order to probe the effect of dust
reddening we consider the @xmath57 colour .
redder colours are more affected by metal line absorption .
we measure the median colour excess @xmath78 as a function of the absorber redshift with the sample divided into rest equivalent width bins , with the results shown in figure [ fig : galex ] .
the global variation can be fit with @xmath79 where @xmath80 , @xmath81 , @xmath82 and the reduced @xmath83 of 0.6 .
we use this reddening measurement to infer the cosmic density of dust in absorbers .
since @xmath84 ( see equation ( [ eq : nhi ] ) below ) , this result is consistent , if marginally , with @xmath85 , an expected result , when we take the fact that the @xmath86 relation has a large scatter into account . from 0.8 to 5.5
[ fig : galex ] this may be taken to be a lower limit on reddening induced by absorber systems taking into account the possibility that some highly obscured quasars might be dropped out of the sample .
budzynski & hewett ( 2011 ) estimated from sdss quasar spectra that about 20% of @xmath87 absorbers are missed due to large obscuration .
the comoving density of a population can be written @xmath88 where @xmath89 is the cross section of the system , @xmath90 is the number intersected in the interval @xmath91 , and the absorbing distance @xmath92 is given by @xmath93 with @xmath94 being the hubble constant at redshift @xmath95 .
the comoving number density of the population is then @xmath96 a similar relation is derived for the cosmic mass density .
we can write the mass density of dust in absorbers as @xmath97 where @xmath98 is the surface dust - mass density of absorbers . in units of the present - day critical density @xmath99 , @xmath100
we note that this estimate of @xmath101 does not require the knowledge of the spatial distribution of absorbers around galaxies , so that we obtain a lower limit on @xmath102 without further knowledge on the distribution of clouds .
the amount of extinction is related to the surface density of dust as , @xmath103 where @xmath104 is the extinction - to - dust - mass coefficient evaluated at the @xmath25 band wavelength , @xmath105 with @xmath106 being the mass density of each species , @xmath107 the extinction cross section of dust in the @xmath25 band per hydrogen atom , @xmath108 the mean molecular weight of gas , and @xmath109 the hydrogen mass . the value of @xmath104 is given in a tabulated form for the model of @xcite at the electronic address quoted above .
for smc type dust we obtain @xmath110 we note that milky way type dust leads to a dust mass larger by a factor of 1.8 at given extinction in the @xmath25-band . , the lower cutoff of our integral . ] the cosmic mass density of dust in absorbers is @xmath111 where the incidence of absorbers is given by @xcite from sdss quasar spectra , @xmath112 with @xmath113 and @xmath114 .
the integrand of eq [ eq : rho_dust ] is shown by the solid line in figure [ fig : nhi_contrib ] .
it shows that absorber systems with @xmath115 , if included , would contribute by 20% to the mass density of dust .
the integral above @xmath116 is about 1% .
the contribution from obscured quasars may also increase the dust amount by 20% if we take literally the effect estimated by budzynski & hewett ( 2011 ) .
+ using equation [ eq : rho_dust ] and the measured reddening shown in figure [ fig : galex ] we estimate the comic density of dust carried by absorbers as a function of redshift .
the results are shown in figure [ fig : omega_dust ] .
we find @xmath117 and an indication for a slight increase from @xmath118 to 0.5 .
we also estimate @xmath119 using equation [ eq : rho_dust ] together with the best fit parameters for eq .
[ eq : e ] .
the result is shown with the dashed curve .
it indicates that @xmath120 with @xmath121 .
we quote as a representative value at a low redshift ( @xmath122 ) , @xmath123 as mentioned above , this provides us with a lower limit on @xmath124 as our analysis does not include contributions from lines of sight that ( i ) produce weak or no absorption , and that ( ii ) intercept high column density clouds with the amount of dust that obscures the background source . we saw that dust missed for the former reason would increase the amount by approximately @xmath125 .
figure [ fig : omega_dust ] includes estimates of the amount of dust in the literature .
msfr give the abundance of dust within the virial radius of representative galaxies at @xmath126 ( @xmath127 ) .
@xcite estimated the dust mass density within the virial radius integrating over all galaxies assuming a typical luminosity function and that the dust amount produced is proportional to luminosity , @xmath128 at the median redshift of the sample they used , and also extended it to include the component outside the virial radius , which is indicated by msfr .
the latter gives @xmath129 , showing that the total abundance of dust is consistent with the amount ought to be produced in star forming activity in cosmic time , when added to dust in disks @xmath130 .
it is noteworthy that the dust amount in clouds is close to that within the virial radius of galaxies , suggesting that dust is predominantly held in clouds within the virial radius .
this is approximately 1/2 the total amount of dust expected to be spread outside galaxies .
we note that the calculation of both @xmath124 and @xmath131 adopts smc type dust and the same reddening - to - dust mass ratio .
the estimate of fukugita & peebles ( 2004 ) refers to dust residing in disks of local galaxies , and driver et al .
( 2007 ) estimate the dust abundance in disc galaxies at @xmath132 , both indicated in the figure for comparison .
this shows that the amount of dust in absorbers is similar to that remaining in galactic discs : much is blown out of galaxies .
the same formalism also applies to estimate the mass density of neutral hydrogen associated with absorbers @xcite .
we have @xmath133 the neutral hydrogen column density of absorbers has been studied by @xcite , who compiled about 200 lyman-@xmath40 measurements of absorbers . using this sample
, @xcite showed that the median @xmath134 of absorbers is described by @xmath135 using this relation clouds below the empirical threshold of star formation in galaxies derived by @xcite . ] and equations [ eq : nestor ] & [ eq : rho_hi ] , we find @xmath136 for the total sample which has median redshift @xmath137
. traced by absorbers . ]
the hi abundance in clouds may slightly decrease towards low redshift , but the change is within the error .
we thus compute the global dust - to - hi ratio for absorbers , @xmath138 including the mass contribution from helium and heavier elements , it gives a dust - to - gas mass ratio of about @xmath139 , which is consistent with the standard value for normal galaxies including the milky way .
let us compare the amount of neutral hydrogen contained in absorbers to the cosmic density .
the hipass hi survey with 1000 galaxies ( zwaan et al .
2003 ) gives @xmath140 from the integration of the hi mass function assuming the schechter form .
the objects with @xmath141 are all identified with optical galaxies at least in high latitude fields where one can avoid confusions .
this suggests that a smooth interpolation of hi objects is also likely to consist of the small galaxy population , rather than another population such as clouds around galaxies , although the possibility is not excluded that it includes contributions from clouds .. ] fukugita & peebles have estimated that , at the present epoch , the total cosmic density of hi amounts to @xmath142 our result shows that absorbers add about a third of the neutral hydrogen amount in galactic discs and bear a fourth of neutral hydrogen amount in the universe .
large quasar samples , aided with precision photometry , have enabled us to study the dust content of absorbers over a broad redshift range . in addition
the combination of high - redshift systems and uv observations allows us to study the short - wavelength extinction properties of their dust grains , all the way to the lyman edge . with minimum assumptions
we have estimated the cosmic density of dust borne by clouds . at low redshift
it amounts to @xmath143 this is comparable to the amount of dust in galactic disks : @xmath144 ( fukugita & peebles 2004 , driver et al .
2007 ) .
fukugita ( 2011 ) estimated the total amount of dust produced in the universe to be @xmath145 this implies that the amount of dust expelled from galaxies is about @xmath146 .
absorbers therefore carry about 1/3 to 1/2 of the total amount of dust expelled by galaxies .
the ratio of the abundances @xmath147 indicates that the gas responsible for absorption should have integrated dust for the corresponding fraction of the cosmic age : clouds should have persisted for a time scale of order several gyr .
the overall gas distribution does not disperse too quickly and lasts _ effectively _ for a significant fraction of the cosmic age . the gas distribution traced by absorbers
should have received dust produced in nearby galaxies and integrated it at least for a few gyr period . in the present study we measured the shape of the extinction curve from @xmath148 ,
at which the absorption of lyman edge contributes predominantly , to longer wavelengths where the opacity is characterized by dust extinction .
the presence of metal absorption lines such as and feii is indicated in the broad - band brightness changes due to the presence of absorbers .
larger samples would allow detections of lyman lines and additional uv metal lines in photometric observations .
we also estimated the hydrogen mass density borne by cloud to be @xmath149 .
this means that the dust - to - hydrogen mass density is of the order of @xmath2 , which is similar to the fraction for normal galaxies , indicating that clouds , thought to be devoid of stars and hence have no sources of dust , can not be aggregates of pristine gas , but are a secondary product from the activity of galaxies .
this supports their outflow origin , which was indicated for star forming galaxies @xcite and the overall redshift dependence of star formation @xcite .
our effective lifetime estimate of the clouds , inferred from the abundance consideration , indicates that such clouds are not ephemeral but last for a long time , consistent with the fact known from early times that absorbers are not only associated with star bursting galaxies but also with galaxies of any morphological types , irrespective of on - going star - forming activity @xcite . wherever stars exist , there was bursting phases in the past , and the cloud can integrate outflows .
the neutral hydrogen abundance in clouds is 5% of the mass density of stars in galaxies , or 10% of gas shed by stars during evolution in the cosmic time , assuming the chabrier initial mass function ( fukugita 2011 ) . for some
actively star - forming galaxy samples it is observed that a significant fraction of , sometimes as much as , the star - forming mass is outflowed ( e.g. , heckman et al .
2000 ; pettini et al .
2002 ; veilleux et al . 2005 ; rupke et al . 2005 ; weiner et al .
our estimate of the neutral hydrogen abundance in absorbers also supports these observations of outflows .
the present analysis is an example of the use of dust as a tracer of mass transactions in the universe with the implication that can be derived by tracing the fate of dust .
is supported by the sloan foundation , the nsf and the henri chrtien grant .
thanks monell foundation at the institute for advanced study , and receives grant - in - aid of the ministry of education in tokyo . | absorbers induce reddening on background quasars .
we measure this effect and infer the cosmic density of dust residing in these systems to be @xmath0 , in units of the critical density of the universe , which is comparable to the amount of dust found in galactic disks or about half the amount inferred to exist outside galaxies .
we also estimate the neutral hydrogen abundance in clouds to be @xmath1 , which is approximately 5% of hydrogen in stars in galaxies . this implies a dust - to - gas mass ratio for clouds of about @xmath2 , which is similar to the value for normal galaxies .
this would support the hypothesis of the outflow origin of clouds , which are intrinsically devoid of stars and hence have no sources of dust .
considerations of the dust abundance imply that the presence of absorbers around galaxies lasts effectively for a few gyr .
high redshift absorbers allow us to measure the rest - frame extinction curve to @xmath3 , at which the absorption by the lyman edge dominates over scattering by dust in the extinction opacity . |
cluster formation in complex fluids is a topic that has attracted considerable attention recently.@xcite the general belief is that a short - range attraction in the pair interaction potential is necessary to initiate aggregation and a long - range repulsive tail in order to limit cluster growth and prevent phase separation .
an alternative scenario for cluster formation pertains to systems whose constituent particles interact by means of purely repulsive potentials .
cluster formation in this case is counterintuitive at first sight : why should particles form clusters if there is no attraction acting between them ?
the answer lies in an additional property of the effective repulsion , namely that of being _ bounded _ , thus allowing full particle overlaps .
though surprising and seemingly unphysical at first , bounded interactions are fully legitimate and natural as effective potentials@xcite between polymeric macromolecular aggregates of low internal monomer concentration , such as polymers,@xcite dendrimers,@xcite microgels,@xcite or coarse - grained block copolymers.@xcite the growing interest in this type of effective interactions is also underlined by the recent mathematical proof of the existence of crystalline ground states for such systems.@xcite cluster formation in the fluid _ and _ in the crystal phases was explicitly seen in the system of penetrable spheres , following early simulation results@xcite and subsequent cell - model calculations.@xcite cluster formation was attributed there to the tendency of particles to create free space by forming full overlaps .
the conditions under which ultrasoft and purely repulsive particles form clusters have been conjectured a few years ago@xcite and explicitly confirmed by computer simulation very recently.@xcite the key lies in the behavior of the fourier transform of the effective interaction potential : for clusters to form , it must contain negative parts , forming thus the class of @xmath1-interactions .
the complementary class of potentials with purely nonnegative fourier transforms , @xmath2 , does not lead to clustering but to remelting at high densities.@xcite an intriguing feature of the crystals formed by @xmath1-systems is the independence of the lattice constant on density,@xcite a feature that reflects the flexibility of soft matter systems in achieving forms of self - organization unknown to atomic ones.@xcite the same characteristic has recently been seen also in slightly modified models that contain a short - range hard core.@xcite in this work , we provide an analytical solution of the crystallization problem and of the properties of the ensuing solids within the framework of an accurate density functional approach .
we explicitly demonstrate the persistence of a single length scale at all densities and for all members of the @xmath1-class and offer thus broad physical insight into the mechanisms driving the stability of the clustered crystals .
we further establish some universal structural properties of all @xmath1-systems both in the fluid and in the solid state , justifying the use of the mean - field density functional on which this work rests .
we make a connection between our results and the harmonic theory of solids in the einstein - approximation .
finally , we establish a connection with suitably - defined infinite dimensional models of hard spheres .
the rest of this paper is organized as follows : in sec .
[ dft : sec ] we derive an accurate density functional by starting with the uniform phase and establishing the behavior of the direct correlation functions of the fluid with density and temperature .
based on this density functional , we perform an analytical calculation of the freezing characteristics of the @xmath1-systems by employing a weak approximation in sec .
[ analytical : sec ] .
the accuracy of this approximation is successfully tested against full numerical minimization of the functional in sec .
[ compare : sec ] . in sec .
[ harmonic : sec ] the equivalence between the density functional and the theory of harmonic crystals is demonstrated , whereas in sec .
[ invpower : sec ] a connection is made with inverse - power potentials . finally , in sec . [ summary : sec ] we summarize and draw our conclusions .
some intermediate , technical results that would interrupt the flow of the text are relegated in the appendix .
in this work , we focus our interest on systems of spherosymmetric particles interacting by means of _ bounded _ pair interactions @xmath3 of the form : @xmath4 with an energy scale @xmath5 and a length scale @xmath6 , and which fulfill the ruelle conditions for stability.@xcite in eq .
( [ gen_poten : eq ] ) above , @xmath7 is some dimensionless function of a dimensionless variable and @xmath8 denotes the distance between the spherosymmetric particles . in the context of soft matter physics , @xmath3 is an effective potential between , e.g. , the centers of mass of macromolecular entities , such as polymer chains or dendrimers.@xcite as the centers of mass of the aggregates can fully overlap without this incurring an infinitely prohibitive cost in ( free ) energy , the condition of boundedness is fulfilled : @xmath9 with some constant @xmath10 . the interaction range is set by @xmath6 , typically the physical size ( e.g. , the gyration radius ) of the macromolecular aggregates that feature @xmath3 as their effective interaction .
in addition to being bounded , the second requirement to be fulfilled by the function @xmath7 is that it decay sufficiently fast to zero as @xmath11 , so that its fourier transform @xmath12 exists . in three spatial dimensions
, we have @xmath13 and , correspondingly , @xmath14 for the fourier transform @xmath15 of the potential , evaluated at wavenumber @xmath16 . our focus in this work is on systems for which @xmath15 is oscillatory , i.e. , @xmath3 features positive and negative fourier components , classifying it as a @xmath1-potential.@xcite though this work is general , for the purposes of demonstration of our results , we consider a particular realization of @xmath1-potentials , namely the generalized exponential model of exponent @xmath17 , ( gem-@xmath17 ) : @xmath18 , \label{gemn : eq}\ ] ] with @xmath19 .
it can be shown that all members of the gem-@xmath17 family with @xmath20 belong to the @xmath1-class , see appendix a. a short account of the freezing and clustering behavior of the gem-4 model has been recently published.@xcite another prominent member of the family is the @xmath21 model , which corresponds to penetrable spheres@xcite with a finite overlap energy penalty @xmath5 . indeed ,
the explicitly calculated phase behavior of these two show strong resemblances , with the phase diagram of both being dominated by the phenomenon of formation of clusters of overlapping particles and the subsequent ordering of the same in periodic crystalline arrangements.@xcite in this work , we provide a generic , accurate , and analytically tractable theory of inhomogeneous phases of @xmath1-systems .
let us start from the simpler system of a homogeneous fluid , consisting of @xmath22 spherosymmetric particles enclosed in a macroscopic volume @xmath23 .
the structure and thermodynamics of the system are determined by the density @xmath24 and the absolute temperature @xmath25 or , better , their dimensionless counterparts : @xmath26 and @xmath27 with boltzmann s constant @xmath28 . as usual
, we also introduce for future convenience the inverse temperature @xmath29 .
we seek for appropriate and accurate closures to the ornstein - zernike relation@xcite @xmath30 connecting the total correlation function @xmath31 to the direct correlation function @xmath32 of the uniform fluid .
one possibility is offered by the hypernetted chain closure ( hnc ) that reads as@xcite @xmath33 - 1 .
\label{hnc : eq}\ ] ] an additional closure , the mean - field approximation ( mfa ) , was also employed and will be discussed later .
our solution by means of approximate closures was accompanied by extensive @xmath34-monte carlo ( mc ) simulations .
we measured the radial distribution function @xmath35 as well as the structure factor @xmath36 , where @xmath37 is the fourier transform of @xmath31 , to provide an assessment of the accuracy of the approximate theories .
we typically simulated ensembles of up to @xmath38 particles for a total of @xmath39 monte carlo steps .
measurements were taken in every tenth step after equilibration .
of the gem-4 model as obtained by monte carlo simulation ( points ) , the hnc - closure ( solid lines ) and the mfa ( dashed lines ) , at various temperatures and densities . for clarity , the curves on every panel have been shifted upwards by certain amounts , which are given below in square brackets , following the value indicating the density @xmath40 .
( a ) @xmath41 and densities , from bottom to top : @xmath42 [ 0 ] ; @xmath43 [ 0.2 ] ; @xmath44 [ 0.4 ] ; @xmath45 [ 0.6 ] ; @xmath46 [ 0.8 ] .
( b ) @xmath47 and densities , from bottom to top : @xmath43 [ 0 ] ; @xmath45 [ 0.2 ] ; @xmath48 [ 0.4 ] ; @xmath49 [ 0.6 ] ; @xmath50 [ 0.8 ] .
( c ) @xmath51 and densities , from bottom to top : @xmath45 [ 0 ] ; @xmath49 [ 0.2 ] ; @xmath52 [ 0.4 ] ; @xmath53 [ 0.6 ] ; @xmath54 [ 0.8].,width=309 ] of the gem-4 model as obtained by monte carlo simulation ( points ) , the hnc - closure ( solid lines ) and the mfa ( dashed lines ) , at various temperatures and densities . for clarity , the curves on every panel have been shifted upwards by certain amounts , which are given below in square brackets , following the value indicating the density @xmath40 .
( a ) @xmath41 and densities , from bottom to top : @xmath42 [ 0 ] ; @xmath43 [ 0.2 ] ; @xmath44 [ 0.4 ] ; @xmath45 [ 0.6 ] ; @xmath46 [ 0.8 ] .
( b ) @xmath47 and densities , from bottom to top : @xmath43 [ 0 ] ; @xmath45 [ 0.2 ] ; @xmath48 [ 0.4 ] ; @xmath49 [ 0.6 ] ; @xmath50 [ 0.8 ] .
( c ) @xmath51 and densities , from bottom to top : @xmath45 [ 0 ] ; @xmath49 [ 0.2 ] ; @xmath52 [ 0.4 ] ; @xmath53 [ 0.6 ] ; @xmath54 [ 0.8].,width=309 ] of the gem-4 model as obtained by monte carlo simulation ( points ) , the hnc - closure ( solid lines ) and the mfa ( dashed lines ) , at various temperatures and densities . for clarity , the curves on every panel have been shifted upwards by certain amounts , which are given below in square brackets , following the value indicating the density @xmath40 .
( a ) @xmath41 and densities , from bottom to top : @xmath42 [ 0 ] ; @xmath43 [ 0.2 ] ; @xmath44 [ 0.4 ] ; @xmath45 [ 0.6 ] ; @xmath46 [ 0.8 ] .
( b ) @xmath47 and densities , from bottom to top : @xmath43 [ 0 ] ; @xmath45 [ 0.2 ] ; @xmath48 [ 0.4 ] ; @xmath49 [ 0.6 ] ; @xmath50 [ 0.8 ] .
( c ) @xmath51 and densities , from bottom to top : @xmath45 [ 0 ] ; @xmath49 [ 0.2 ] ; @xmath52 [ 0.4 ] ; @xmath53 [ 0.6 ] ; @xmath54 [ 0.8].,width=309 ] in fig . [ gofr : fig ] we show comparisons for the function @xmath55 as obtained from the mc simulations and from the hnc closure for a variety of temperatures and densities .
it can be seen that agreement between the two is obtained , to a degree of quality that is excellent .
tiny deviations between the hnc and mc results appear only at the highest density and for a small region around @xmath56 for low temperatures , @xmath41 .
otherwise , the system at hand is described by the hnc with an extremely high accuracy and for a very broad range of temperatures and densities .
we note that , although in fig .
[ gofr : fig ] we restrict ourselves to temperatures @xmath57 , the quality of the hnc remains unaffected also at higher temperatures.@xcite in attempting to gain some insight into the remarkable ability of the hnc to describe the fluid structure at such a high level of accuracy , it is useful to recast this closure in density - functional language .
following the famous percus idea,@xcite the quantity @xmath58 can be identified with the nonuniform density @xmath59 of an inhomogeneous fluid that results when a single particle is kept fixed at the origin , exerting an ` external ' potential @xmath60 on the rest of the system . following standard procedures from density functional theory , we find that the sought - for density profile @xmath59 is given by @xmath61}{\delta \rho(r)}\right\ } , \label{profile_dft : eq}\ ] ] where @xmath62 is the thermal de broglie wavelength and @xmath63 the chemical potential associated with average density @xmath64 and temperature @xmath25 .
moreover , @xmath65 $ ] is the intrinsic excess free energy , a _
unique _ functional of the density @xmath59 . as such , @xmath65 $ ] can be expanded in a functional taylor series around its value for a uniform liquid of some ( arbitrary ) reference density @xmath66 . for the problem at hand
, @xmath67 is a natural choice and we obtain@xcite @xmath68 = \beta f_{\rm ex}(\rho ) & - & \sum_{n=1}^{\infty}\frac{1}{n!}\int\int\cdots\int { \rm d}^3r_1{\rm d}^3r_2\ldots{\rm d}^3r_n \\ \nonumber & \times & c_0^{(n)}({\bf r}_1,{\bf r}_2,\ldots,{\bf r}_n;\rho ) \\ & \times & \delta\rho({\bf r}_1)\delta\rho({\bf r}_2)\ldots\delta\rho({\bf r}_n ) , \label{taylor_expand : eq}\end{aligned}\ ] ] where @xmath69 denotes the excess free energy of the _ homogeneous _ fluid , as opposed to that of the inhomogeneous fluid , @xmath65 $ ] , and @xmath70 the basis of the expansion given by eq .
( [ taylor_expand : eq ] ) above is the fact that @xmath65 $ ] is the generating functional for the hierarchy of the direct correlation functions ( dcf s ) @xmath71 . in particular
, @xmath72 is the @xmath73-th functional derivative of the excess free energy with respect to the density field , evaluated at the uniform density @xmath64:@xcite @xmath74 } { \delta\rho({\bf r}_1)\delta\rho({\bf r}_2)\ldots \delta\rho({\bf r}_n)}\bigg|_{\rho ( . )
= \rho}. \label{cofrn : eq}\ ] ] as the functional derivatives are evaluated for a uniform system , translational and rotational invariance reduce the number of variables on which the @xmath73-th order dcf s @xmath71 depend ; in fact , @xmath75 is a position - independent constant and equals @xmath76 , where @xmath77 is the excess chemical potential.@xcite similarly , @xmath78 is simply the ornstein - zernike direct correlation function @xmath79 entering in eqs .
( [ oz : eq ] ) and ( [ hnc : eq ] ) above .
the hnc closure is equivalent to jointly solving eqs .
( [ oz : eq ] ) and ( [ profile_dft : eq ] ) by employing an approximate excess free energy functional @xmath65 $ ] , arising by a truncation of the expansion of eq .
( [ taylor_expand : eq ] ) at @xmath80 , i.e. , discarding all terms with @xmath81 ; this is the famous ramakrishnan - yussouff second - order approximation.@xcite indeed , the so - called bridge function @xmath82 can be written as an expansion over integrals involving as kernels all the @xmath71 with @xmath81 and the hnc amounts to setting the bridge function equal to zero.@xcite whereas in many cases , such as the one - component plasma,@xcite and other systems with long - range interactions , the hnc is simply an adequate or , at best , a very good approximation , for the case at hand the degree of agreement between the hnc and simulation is indeed extremely high .
what is particularly important is that the accuracy of the hnc persists for a very wide range of densities , at all temperatures considered .
this fact has far - reaching consequences , because it means that the corresponding profiles @xmath83 that enter the multiple integrals in eq .
( [ taylor_expand : eq ] ) vary enormously depending on the uniform density considered .
thus , it is tempting to conjecture that for the systems under consideration ( soft , penetrable particles at @xmath84 ) , not simply the integrals with @xmath81 vanish but rather the kernels themselves . in other words ,
@xmath85 calculated in the hnc and its mfa - approximation , @xmath86 , for a gem-4 model , at various densities indicated in the legends .
the insets show the mfa approximation for the same quantity , which is density - independent .
each panel corresponds to a different temperature : ( a ) @xmath41 ; ( b ) @xmath47 ; ( c ) @xmath51 .
these are exactly the same parameter combinations as the ones for which @xmath55 is shown in fig .
[ gofr : fig].,width=309 ] calculated in the hnc and its mfa - approximation , @xmath86 , for a gem-4 model , at various densities indicated in the legends .
the insets show the mfa approximation for the same quantity , which is density - independent .
each panel corresponds to a different temperature : ( a ) @xmath41 ; ( b ) @xmath47 ; ( c ) @xmath51 .
these are exactly the same parameter combinations as the ones for which @xmath55 is shown in fig .
[ gofr : fig].,width=309 ] calculated in the hnc and its mfa - approximation , @xmath86 , for a gem-4 model , at various densities indicated in the legends .
the insets show the mfa approximation for the same quantity , which is density - independent .
each panel corresponds to a different temperature : ( a ) @xmath41 ; ( b ) @xmath47 ; ( c ) @xmath51 .
these are exactly the same parameter combinations as the ones for which @xmath55 is shown in fig .
[ gofr : fig].,width=309 ] the behavior of the higher - order dcf s is related to the density - derivative of lower - order ones through certain sum rules , to be discussed below .
hence , it is pertinent to examine the density dependence of the dcf @xmath32 of the hnc . in fig .
[ cofr : fig ] we show the difference between the dcf @xmath87 and the mean - field approximation ( mfa ) to the same quantity : @xmath88 eq .
( [ cmfa : eq ] ) above is _ meaningless _ if the pair potential @xmath3 diverges as @xmath89 , because @xmath32 has to remain finite at all @xmath8 , as follows from exact diagrammatic expansions of the same.@xcite in fact , the form @xmath90 denotes the large-@xmath8 asymptotic behavior of @xmath32 . in our case , however , where @xmath3 lacks a hard core , the mfa - form for @xmath32 can not be a priori rejected on fundamental grounds ; the quantity @xmath91 remains bounded as @xmath92 .
in fact , as can be seen in fig .
[ cofr : fig ] , the deviations between the mfa and the hnc - closure are very small for @xmath93 . in addition , the differences between @xmath87 and @xmath94 drop , both in absolute and in relative terms , as temperature grows , see the trend in figs .
[ cofr : fig](a)-(c ) . at fixed temperature ,
the evolution of the difference with density is nonmonotonic : it first drops as density grows and then it starts growing again at the highest densities shown in the three panels of fig .
[ cofr : fig ] . motivated by these findings
, we employ now a second closure , namely the above - mentioned mfa , eq .
( [ cmfa : eq ] ) . introducing the latter into the ornstein - zernike relation , eq .
( [ oz : eq ] ) , we obtain the mfa - results for the radial distribution function @xmath55 that are also shown in fig .
[ gofr : fig ] with dashed lines . before proceeding to a critical comparison between the @xmath55 obtained from the two closures , it is useful to make a clear connection between the mfa and the hnc .
as mentioned above , the members of the sequence of the @xmath73-th order direct correlation functions are not independent from one another ; rather , they are constrained to satisfy a corresponding hierarchy of sum rules , namely:@xcite @xmath95 in particular , for @xmath96 , we have @xmath97 where we have used the translational and rotational invariance of the fluid phase to reduce the number of arguments of the dcf s and we show explicitly the generic dependence of @xmath32 on @xmath64 . in the mfa , one assumes @xmath86 , with the immediate consequence @xmath98 eqs .
( [ sumrule : eq ] ) and ( [ deriv : eq ] ) imply that the integral of @xmath99 with respect to any of its arguments must _ vanish _ for _ arbitrary _ density .
as @xmath99 has a complex dependence on its arguments , this is a strong indication that @xmath99 itself vanishes .
in fact , both for the barrat - hansen - pastore factorization approximation for this quantity@xcite and for the alternative , denton - ashcroft @xmath16-space factorization of the same,@xcite the vanishing of the right - hand side of eq .
( [ sumrule : eq ] ) implies that @xmath100 .
now , if @xmath99 vanishes , so does also its density derivative and use of sum rule ( [ inductive : eq ] ) for @xmath101 implies @xmath102 .
successive use of the same for higher @xmath73-values leads then to the conclusion that in the mfa : @xmath103 we can now see that the accuracy of the hnc stems from the fact that for these systems we can write @xmath104 where @xmath105 is a small function at all densities @xmath64 , offering concomitantly a very small , and _ the only _ , contribution to the quantity @xmath106 . this implies that @xmath99 itself is negligible by means of eq .
( [ sumrule : eq ] ) . repeated use of eq .
( [ inductive : eq ] ) leads then to eqs .
( [ approx : eq ] ) and shows that the contributions from the @xmath107 terms , that are ignored in the hnc , are indeed negligible .
the hnc is , thus , very accurate , due to the strong mean - field character of the fluids at hand.@xcite the deviations between the hnc and the mfa come through the function @xmath105 above .
let us now return to the discussion of the results for @xmath55 and @xmath32 and the relative quality of the two closures at various thermodynamic points . referring first to fig .
[ gofr : fig](c ) , we see that at @xmath51 both the hnc and the mfa perform equally well .
the agreement between the two ( and between the mfa and mc ) worsens somewhat at @xmath47 and even more at @xmath41 .
the mfa is , thus , an approximation valid for @xmath84 , in agreement with previous results.@xcite the reason lies in the accuracy of the low - density limit of the mfa . in general , as @xmath108 , @xmath32 tends to the mayer function @xmath109 - 1 $ ] .
if @xmath84 , one may expand the exponential to linear order and obtain @xmath110 , so that the mfa can be fulfilled . in the hnc , it is implicitly assumed that all dcf s with @xmath81 vanish . in the mfa , this is also the case .
the two closures differ in one important point , though : in the hnc , the second - order direct correlation function is _ not _ prescribed but rather _ determined _ , so that both the ` test - particle equation ' , eq . ( [ profile_dft : eq ] ) , and the ornstein - zernike relation , eq .
( [ oz : eq ] ) , are fulfilled . in the mfa , it is a priori assumed that @xmath86 , which is introduced into the ornstein - zernike relation and thus @xmath31 is determined .
this is one particular way of obtaining @xmath55 in the mfa , called the ornstein - zernike route .
alternatively , one could follow the test - particle route in solving eq .
( [ profile_dft : eq ] ) in conjunction with eq .
( [ taylor_expand : eq ] ) and the mfa - approximation , eq .
( [ cmfa : eq ] ) . in this case , the resulting expression for the total correlation function @xmath31 in the mfa reads as @xmath111 - 1 , \label{testpart : eq}\ ] ] with @xmath112 denoting the convolution .
previous studies with ultrasoft systems have shown that the test - particle @xmath31 from the mfa is closer to the hnc - result than the mfa result obtained from the ornstein - zernike route.@xcite the discrepancy between the two is a measure of the approximate character of the mfa ; were the theory to be exact , all routes would give the same result .
as a way to quantify the approximations involved in the mfa , let us attempt to _ impose _ consistency between the test - particle and ornstein - zernike routes .
since @xmath86 in this closure , the exponent in eq .
( [ testpart : eq ] ) above is just the right - hand side of the ornstein - zernike relation , eq .
( [ oz : eq ] ) .
thus , if we insist that the latter is fulfilled , we obtain the constraint @xmath113 - 1 , \label{hofr : eq}\ ] ] which is strictly satisfied only for @xmath114 . however , as long as @xmath115 , one can linearize the exponential and an identity follows ; the internal inconsistency of the mfa is of quadratic order in @xmath31 and it follows that the mfa provides an accurate closure for the systems at hand , as long as @xmath116 remains small .
this explains the deviations between mfa and mc seen at small @xmath8 for the highest density at @xmath51 , fig .
[ gofr : fig](c ) . the same effect
can also be seen in fig .
[ cofr : fig](c ) as a growth of the discrepancy between @xmath87 and @xmath94 at small @xmath8-values for the highest density shown . in absolute terms , however , this discrepancy remains very small .
note also that discrepancies at small @xmath8-values become strongly suppressed upon taking a fourier transform , due to the additional geometrical @xmath117-factor involved in the three - dimensional integration .
it can therefore be seen that the mfa and the hnc are closely related to one another : the hnc is so successful due to the strong mean - field character of the systems under consideration .
this fact has also been established and extensively discussed for the case of the gaussian model,@xcite i.e. , the @xmath118 member of the gem-@xmath17 class .
once more , the hnc and the mfa there are very accurate for high densities and/or temperatures , where @xmath119 and the system s behavior develops similarities with an ` incompressible ideal gas',@xcite in full agreement with the remarks presented above .
subsequently , the mfa and hnc closures have been also successfully applied to the study of structure and thermodynamics of binary soft mixtures.@xcite of the gem-4 model at temperature @xmath47 and various densities , indicated below . for clarity , the curves
have been shifted vertically by amounts shown in the square brackets , following the numbers that indicate the values of the density @xmath40 . from bottom to top : @xmath43 [ 0 ] ; @xmath45 [ 0.5 ] ; @xmath48 [ 1.0 ] ; @xmath49 [ 1.5 ] ; @xmath50 [ 2.0 ] .
the points are results from monte carlo simulations and the dashed lines from the hnc .
as the hnc- and mfa - curves run very close to each other , we show the mfa - result by the solid curve only for the highest density , @xmath50.,width=321 ] a crucial difference between the gaussian model , which belongs to the @xmath2-class , and members of the @xmath1-class , which are the subject of the present work , lies in the consequences of the mfa - closure on the structure factor @xmath120 of the system .
since @xmath86 , the ornstein - zernike relation leads to the expression @xmath121 whereas for @xmath2-potentials @xmath120 is devoid of pronounced peaks that exceed the asymptotic value @xmath122 , for @xmath1-systems a local maximum of @xmath120 appears at the value @xmath123 for which @xmath15 attains its negative minimum . in fig .
[ sofk : fig ] we show representative results for the system at hand , where it can also be seen that the hnc and the mfa yield practically indistinguishable results . in full agreement with the mc simulations ,
the location of the main peak of @xmath120 is _ density - independent _ , a feature unknown for usual fluids , having its origin in the fact that @xmath32 itself is density independent .
associated with this is the development of a @xmath124-line,@xcite also known as _
kirkwood instability_,@xcite on which the denominator in eq .
( [ sofk : eq ] ) vanishes at @xmath123 and thus @xmath125 .
the locus of points @xmath126 on the density - temperature plane for which this divergence takes place is , evidently , given by @xmath127 in the region @xmath128 ( equivalently : @xmath129 ) on the @xmath130-plane , the mfa predicts that the fluid is absolutely unstable , since the structure factor there has multiple divergences and also develops negative parts .
this holds _ only _ for @xmath1-systems ; for @xmath2-ones the very same line of argumentation leads to the opposite conclusion , namely that the fluid is the phase of stability at high densities and/or temperatures .
the latter conclusion has already been reached by stillinger and coworkers@xcite in their pioneering work of the gaussian model in the mid-1970 s , and explicitly confirmed by extensive theory and computer simulations many years later.@xcite however , stillinger s original argument was based on duality relations that are strictly fulfilled only for the gaussian model , whereas the mfa - arguments are quite general .
having established the validity of the mfa for vast domains in the phase diagram of the systems under consideration as far as the _ uniform _ fluids are concerned , we now turn our attention to nonuniform ones .
apart from an obvious general interest in the properties of nonuniform fluids , the necessity to consider deviations from homogeneity in the density for @xmath1-models is dictated by the @xmath124-instability mentioned above : the theory of the uniform fluid contains its own breakdown , thus the system has to undergo a phase transformation to a phase with a spontaneously broken translational symmetry . whether this transformation takes place _ exactly _ on the instability line or already at densities @xmath131 ( or temperatures @xmath132 ) and which is the stable phase are some of the questions that have to be addressed .
density - functional theory of inhomogeneous systems is the appropriate theoretical tool in this direction .
let us consider a path @xmath133 in the space of density functions , which is characterized by a single parameter @xmath134 ; this path starts at some reference density @xmath135 and terminates at another density @xmath136 .
the uniqueness of the excess free energy functional and its dependence on the inhomogeneous density field @xmath136 allow us to integrate @xmath137/\partial\chi$ ] along this path , obtaining @xmath65 $ ] , provided that @xmath138 $ ] is known .
a convenient parametrization reads as @xmath139 $ ] , with @xmath140 corresponding to @xmath135 and @xmath141 to @xmath136 .
the excess free energy of the final state can be expressed as@xcite @xmath68 & = & \beta f_{\rm ex}[\rho_{\rm r } ] \\ & - & \int_0^{1 } { \rm d}\chi \int{\rm d}^3r c^{(1)}({\bf r};[\rho_{\chi } ] ) \delta\rho({\bf r } ) , \label{chi_path : eq}\end{aligned}\ ] ] where @xmath142)$ ] denotes the first functional derivative of the quantity @xmath143 $ ] evaluated at the inhomogeneous density @xmath133 , and @xmath144 . since @xmath142)$ ] is in its own turn a unique functional of the density profile , repeated use of the same argument leads to a functional taylor expansion of the excess free energy around that of a reference system , an expansion that extends to infinite order .
for the _ particular _ choice of a _ uniform _ reference system , @xmath145 , we obtain then the taylor series of eq .
( [ taylor_expand : eq ] ) . in general , however , the reference system does not have to have the same average density as the final one , hence the uniform density @xmath64 there must be replaced by a more general quantity @xmath66 . the usefulness of eq .
( [ taylor_expand : eq ] ) in calculating the free energies of extremely nonuniform phases , such as crystals , is limited both on principal and on practical grounds .
fundamentally , there is _ no small parameter _
guiding such an expansion , since the differences between the nonuniform density of a crystal and that of a fluid are enormous ; the former has extreme variations between lattice- and interstitial regions .
hence , the very convergence of the series is in doubt.@xcite in practice , the direct correlation functions for @xmath146 are very cumbersome to calculate@xcite and those for @xmath147 are practically unknown.@xcite the solution is either to arbitrarily terminate the series at second order@xcite or to seek for nonperturbative functionals.@xcite in our case , however , things are different because , for the systems we consider , we have given evidence that the dcf s of order @xmath107 are extremely small and we take them at this point as vanishing . then , the functional taylor expansion of the free energy @xmath65 $ ] terminates ( to the extent that the approximation holds ) at second order .
the taylor series becomes a finite sum and convergence is not an issue any more .
let us , accordingly , expand @xmath65 $ ] around an _ arbitrary _ , homogeneous reference fluid of density @xmath148 , taking into account that the volume @xmath23 is fixed but the system with density @xmath136 contains @xmath22 particles , whereas the reference fluid contains @xmath149 particles and , in general , @xmath150 : @xmath68 & = & \beta f_{\rm ex}(\rho_0 ) - c_0^{(1)}(\rho_0)\int{\rm d}^3r[\rho({\bf r } ) - \rho_0 ] \\ \nonumber & - & \frac{1}{2}\int\int{\rm d}^3r{\rm d}^3r ' c_0^{(2)}(|{\bf r}-{\bf r'}|;\rho_0 ) \\ & \times & [ \rho({\bf r } ) - \rho_0][\rho({\bf r ' } ) - \rho_0 ] .
\label{taylor : eq}\end{aligned}\ ] ] using @xmath151 and the sum rule ( [ inductive : eq ] ) for @xmath152 together with the vanishing of the excess chemical potential at zero density , we readily obtain @xmath153 formally substituting in eq .
( [ taylor : eq ] ) , @xmath154 and @xmath155 and making use of the fact that the excess free energy of a system vanishes with the density , we also obtain the dependence of the fluid excess free energy on the density : @xmath156 with the particle number @xmath149 _ of the reference fluid _ and the fourier transform @xmath157 of the interaction potential . introducing eqs .
( [ c1:eq ] ) and ( [ frex_liq : eq ] ) into the taylor expansion , eq .
( [ taylor : eq ] ) above , we obtain : @xmath68 & = & \frac{n_0}{2}\beta{\tilde v}(0)\rho_0 \\
\nonumber & + & \beta{\tilde v}(0)\rho_0 ( n - n_0 ) \\ \nonumber & + & \frac{\beta}{2}\int\int{\rm d}^3r{\rm d}^3r'v(|{\bf r } - { \bf r'}| ) \rho({\bf r})\rho({\bf r ' } ) \\ \nonumber & - & \beta\rho_0\int\int{\rm d}^3r{\rm d}^3r'v(|{\bf r } - { \bf r'}| ) \rho({\bf r } ) \\ & + & \frac{n_0}{2}\beta{\tilde v}(0)\rho_0 .
\label{expand : eq}\end{aligned}\ ] ] introducing @xmath158 the fourth term above becomes : @xmath159 now the sum of the 1st , 2nd , 4th and 5th term in eq .
( [ expand : eq ] ) yields : @xmath160 \\ = \beta{\tilde v}(0)\left[n_0\rho_0 + \rho_0(n - n_0 ) - n\rho_0\right ] = 0.\end{aligned}\ ] ] this is a remarkable cancelation because then only the 3rd term in eq .
( [ expand : eq ] ) survives and we obtain : @xmath161 = \frac{1}{2}\int\int{\rm d}^3r{\rm d}^3r ' v(|{\bf r } - { \bf r'}|)\rho({\bf r})\rho({\bf r ' } ) , \label{dftmfa : eq}\ ] ] which is our desired result.@xcite the derivation above demonstrates that the excess free energy of _ any _ inhomogeneous phase for our ultrasoft fluids is given by eq .
( [ dftmfa : eq ] ) , irrespectively of the density of the reference fluid @xmath66 .
this is particularly important because , usually , functional taylor expansions are carried out around a reference fluid whose density lies close to the average density of the inhomogeneous system ( crystal , in our case ) .
however , in our systems this is impossible .
the crystals occur predominantly in domains of the phase diagram in which the reference fluid is meaningless , because they are on the high density side of the @xmath124-line .
it is therefore important to be able to justify the use of the functional and to avoid the inherent contradiction of expanding around an unstable fluid . in practice , of course ,
the higher - order dcf s do not exactly vanish , hence deviations from result ( [ dftmfa : eq ] ) are expected to occur , in particular at low temperatures and densities . nevertheless , the comparisons with simulations , e.g. , in refs .
[ ] , [ ] and [ ] fully justify our approximation .
a mathematical proof of the mean - field character for fluids with infinitely long - range and infinitesimally strong repulsions has existed since the late 1970 s , see refs . [ ] and [ ] . however , even far away from fulfillment of this limit , and for conditions that are quite realistic for soft matter systems , the mean - field behavior continues to be valid.@xcite the mean - field result of eq .
( [ dftmfa : eq ] ) has been put forward for the gaussian model at high densities,@xcite on the basis of physical argumentation : in the absence of diverging excluded - volume interactions , at sufficiently high densities any given particle sees an ocean of others the classical mean - field picture .
the mean - field character of the gaussian model for moderate to high temperatures was demonstrated independently in ref .
, we have provided a more rigorous justification of its validity , based on the vanishing of high - order direct correlation functions in the fluid .
it must also be noted that the mean - field approximation has recently been applied to a system with a broad shoulder and a much shorter hard - core interaction , providing good agreement with simulation results@xcite and allowing for the formulation of a generalized clustering criterion for the inhomogeneous phases .
an astonishing similarity exists between the mean - field functional of eq .
( [ dftmfa : eq ] ) and an exact result derived for infinite - dimensional hard spheres .
indeed , for this case frisch _
et al._@xcite as well as bagchi and rice@xcite have shown that @xmath162 = -\frac{1}{2}\int\int{\rm d}^d r{\rm d}^d r ' f(|{\bf r } - { \bf r'}|)\rho({\bf r})\rho({\bf r ' } ) , \label{mayer : eq}\ ] ] where @xmath163 and @xmath164 is the mayer function of the infinite dimensional hard spheres .
again , one has a bilinear excess functional whose integration kernel does not depend on the density ; in this case , this is minus the ( bounded ) mayer function whereas for mean - field fluids , it is the interaction potential itself , divided by the thermal energy @xmath165 .
in fact , the mayer function and the direct correlation function coincide for infinite - dimensional systems and higher - order contributions vanish there as well,@xcite making the analogy with our three - dimensional , ultrasoft systems complete .
accordingly , infinite dimensional hard spheres have an instability at some finite @xmath16 at the density @xmath166 , given by @xmath167 .
this so - called _
kirkwood instability_@xcite is of the same nature as our @xmath124-line but hard hyperspheres are athermal , so it occurs at a single point on the density axis and not a line on the density - temperature plane .
following kirkwood s work,@xcite it was therefore argued@xcite that hard hyperspheres might have a second - order freezing transition at the density @xmath166 expressed as @xmath168d^{1/6 } , \label{rhostar : eq}\ ] ] where the limit @xmath163 must be taken and @xmath169 is the value of the minimum of the bessel function @xmath170 as @xmath163 . note that @xmath171 as @xmath163 .
later on , frisch and percus argued that most likely the kirkwood instability is never encountered because it is preempted by a first - order freezing transition.@xcite in what follows , we will show analytically that this is also the case for our systems , which might provide a finite - dimensional realization of the above - mentioned mathematical limit .
as mentioned above , an obvious candidate for a spatially modulated phase is a periodic crystal .
the purpose of this section is to employ density functional theory in order to calculate the freezing properties . under some weak , simplifying assumptions ,
the problem can be solved analytically . adding the ideal contribution to the excess functional of eq .
( [ dftmfa : eq ] ) , the free energy of any spatially modulated phase is obtained as @xmath172 & = & f_{\rm id}[\rho ] + f_{\rm ex}[\rho ] \\
\nonumber & = & k_{\rm b}t\int{\rm d}^3r\left[\ln\left[\rho({\bf r})\lambda^3\right ] -1\right ] \\ & + & \frac{1}{2}\int\int{\rm d}^3r{\rm d}^3r ' v(|{\bf r}-{\bf r'}|)\rho({\bf r})\rho({\bf r ' } ) .
\label{totalfren : eq}\end{aligned}\ ] ] as we are interested in crystalline phases , we parametrize the density profile as a sum of gaussians centered around the lattice sites @xmath173 , forming a bravais lattice . in sharp contrast with systems interacting by hard , diverging potentials , however , the assumption of one particle per lattice site has to be dropped .
indeed , it will be seen that @xmath1 systems employ the strategy of optimizing their lattice constant by adjusting the number of particles per lattice site , @xmath174 , at any given density @xmath64 and temperature @xmath25 .
accordingly , we normalize the profiles to @xmath174 and write @xmath175 where the occupation variable @xmath174 and the localization parameter @xmath176 have to be determined variationally , and the lattice site density @xmath177 is expressed as @xmath178 contrary to crystals of single occupancy , thus , the number of particles @xmath22 and the number of sites @xmath179 of the bravais lattice do not coincide . in particular , it holds @xmath180 and we are interested in multiple site occupancies , i.e. , @xmath181 or even @xmath182 ; it will be shown that this clustering scenario indeed minimizes the crystal s free energy .
it is advantageous , at this point , to express the periodic density profile of eq .
( [ gaussian_real : eq ] ) as a fourier series , introducing the fourier components @xmath183 of the same : @xmath184 where the set @xmath185 contains all reciprocal lattice vectors ( rlvs ) of the bravais lattice formed by the set @xmath173 .
accordingly , the inverse of ( [ denfour : eq ] ) reads as:@xcite @xmath186 where the first integral extends over the elementary unit cell @xmath187 of the crystal and @xmath188 is the volume of @xmath187 , containing a single lattice site .
the second integral extends over all space , where use of the periodicity of @xmath136 and its expression as a sum over lattice site densities , eq .
( [ gaussian_real : eq ] ) , has been made .
using eq .
( [ orbital : eq ] ) we obtain @xmath189 note that the site occupancy @xmath174 does not appear explicitly in the functional form of the fourier components of @xmath136 , a feature that may seem paradoxical at first sight . however , for fixed density @xmath64 and any crystal type , the lattice constant and thus also the reciprocal lattice vectors @xmath190 are affected by the possibility of clustering , thus the dependence on @xmath174 remains , albeit in an implicit fashion . with the density being expressed in reciprocal space , the excess free energy takes a simple form that reads as @xmath191 the ideal term , @xmath192 $ ] , can also be approximated analytically , provided that the gaussians centered at different lattice sites do not overlap .
let @xmath193 denote the lattice constant of any particular crystal .
then , for @xmath194 , the ideal free energy of the crystal takes the form @xmath195 where the trivial last term will be dropped in what follows , since is also appears in the expression of the free energy of the fluid and does not affect any phase boundaries .
putting together eqs .
( [ fexk : eq ] ) and ( [ fida : eq ] ) , we obtain a variational free energy per particle , @xmath196 , for the crystal , that reads as @xmath197 \\ & + & \frac{\rho^*}{2}\sum_{\bf y}\tilde\phi(y ) e^{-y^2/(2\alpha^ * ) } , \label{fren_var : eq}\end{aligned}\ ] ] where @xmath198 and @xmath199 . in the list of arguments of @xmath196
the first two are variational parameters whereas the last two denote simply its dependence on temperature and density .
the free energy per particle , @xmath200 of the crystal is obtained by minimization of @xmath196 , i.e. , @xmath201 in carrying out the minimization , it proves useful to measure the localization length of the gaussian profile , @xmath202 , in units of the lattice constant @xmath193 instead of units of @xmath6 . to perform this change
, we first express the average density @xmath64 of the crystal in terms of @xmath174 and @xmath193 as @xmath203 where @xmath204 is a lattice - dependent numerical coefficient of order unity .
introduce now the quantity @xmath205 .
using eq .
( [ den_a : eq ] ) above , we obtain @xmath206 this change of variables is just a mathematical transformation that simplifies the mathematics to follow ; all results to be derived maintain their validity also in the original representation . for a further discussion of this point , see also appendix b. next we make the simplifying approximation to ignore in the sum over reciprocal lattice vectors on the right - hand - side of eq .
( [ fren_var : eq ] ) above all the rlvs beyond the first shell , whose length is @xmath207 .
this is justified already because of the exponentially damping factors @xmath208 $ ] in the sum .
in addition , the coefficients @xmath209 themselves decay to zero as @xmath210 , with an asymptotic behavior that depends on the form of @xmath7 in real space . the length of the first shell of rlvs of any bravais lattice of lattice constant @xmath193 scales as @xmath211 , with some positive , lattice - dependent numerical constant @xmath212 of order unity . together with eqs .
( [ den_a : eq ] ) this implies that the length of the first rlv depends on the aggregation number @xmath174 as @xmath213 and using eq .
( [ amin_nc : eq ] ) , we see that the ratio @xmath214 takes a form that depends _ solely _ on the parameter @xmath215 , namely @xmath216 introducing eqs .
( [ amin_nc : eq ] ) and ( [ d : eq ] ) into ( [ fren_var : eq ] ) , we obtain another functional form for the variational free energy , @xmath217 , expressed in the new variables .
it can be seen that upon making the transformation ( [ amin_nc : eq ] ) , the term @xmath218 $ ] delivers a contribution _ minus _ @xmath219 that exactly cancels the same term with a positive sign on the right - hand side of eq .
( [ fren_var : eq ] ) .
accordingly , the _ only _ remaining quantity of the variational free energy that still depends on @xmath174 is the length of the first nonvanishing rlv , @xmath220 , whose @xmath174 dependence is expressed by eq .
( [ y1nc : eq ] ) above .
putting everything together , we obtain @xmath221 - \ln z \right ] \\ \nonumber & + & \frac{1}{2}\rho^*\tilde\phi(0 ) \\ & + & \frac{\xi_1\rho^*}{2}\tilde\phi\left(y_1(n_c)\right)e^{-\gamma\zeta^2/2 } , \label{fbar : eq}\end{aligned}\ ] ] where @xmath222 is the coordination number of the _ reciprocal _ lattice .
minimizing @xmath223 with respect to @xmath174 is trivial and using eq .
( [ d : eq ] ) we obtain @xmath224 where the prime denotes the derivative with respect to the argument .
evidently , @xmath220 coincides with @xmath225 , the dimensionless wavenumber for which the dimensionless fourier transform of the interaction potential attains its negative minimum .
the other mathematical solution of ( [ partialnc : eq ] ) , @xmath226 , can be rejected because it yields nonpositive second derivatives or , on physical grounds , because it corresponds to a crystal with @xmath227 , whose occurrence would violate the thermodynamic stability of the system . regarding second derivatives , it can be easily shown that @xmath228 and @xmath229 irrespective of @xmath215 .
having shown the coincidence of @xmath220 with @xmath230 , we set @xmath231 in eq .
( [ fbar : eq ] ) above .
further , we notice that the term @xmath232 $ ] on the right - hand side of eq .
( [ fbar : eq ] ) gives the ideal free energy of a uniform fluid of density @xmath233 and the term @xmath234 the excess part of the same , see eq .
( [ frex_liq : eq ] ) .
subtracting , thus , the total fluid free energy per particle , @xmath235 , we introduce the difference @xmath236 , which reads as @xmath237 \\ & + & \frac{\xi_1\rho^*}{2}\tilde\phi(y_*)e^{-\gamma\zeta^2/2}. \label{deltaf : eq}\end{aligned}\ ] ] the requirement of no overlap between gaussians centered on different lattice sites restricts @xmath215 to be small ; a very generous upper limit is @xmath238 . for such small values of @xmath215 , the first term on the right - hand side of eq .
( [ deltaf : eq ] ) above is positive .
this positivity expresses the entropic cost of localization that a crystal pays , compared to the fluid in which the delocalized particles possess translational entropy .
this cost must be compensated by a gain in the excess term , which is only possible if @xmath239 .
an additional degree of freedom is offered by the candidate crystal structures .
the excess free energy is minimized by the direct bravais lattice whose reciprocal lattice has the maximum possible coordination number @xmath222 .
the most highly coordinated periodic arrangement of sites is fcc , for which @xmath240 .
therefore , in the framework of this approximation , the stable lattice is bcc . it must be emphasized , though , that these results hold as long as only the first shell of rlvs is kept in the excess free energy .
inclusion of higher - order shells can , under suitable thermodynamic conditions , stabilize fcc in favor of bcc .
we will return to this point later .
choosing now @xmath193 as the edge - length of the _ conventional _ lattice cell of the bcc - lattice , we have @xmath241 and @xmath242 . evidently , the lattice constant of the crystal is density - independent , @xmath243 , contrary to the case of usual crystals , for which @xmath244 .
the density - independence of @xmath193 is achieved by the creation of clusters that consist of @xmath174 particles , each of them occupying a lattice site .
the proportionality relation connecting @xmath174 and @xmath233 follows from eq .
( [ y1nc : eq ] ) and reads as @xmath245 it remains to minimize @xmath223 ( equivalently , @xmath246 ) with respect to @xmath215 to determine the free energy of the crystal .
we are interested , in particular , in estimating the ` freezing line ' , determined by the equality of free energies of the fluid and the solid.@xcite accordingly , we search for the simultaneous solution of the equations @xmath247 resulting into @xmath248 and @xmath249 .
\label{dfzero : eq}\ ] ] substituting ( [ dfzero : eq ] ) into ( [ dfdg : eq ] ) and using @xmath250 and @xmath242 , we obtain an implicit equation for @xmath215 that reads as @xmath251 , \label{gamma : eq}\ ] ] and has two solutions , @xmath252 and @xmath253 . due to ( [ second_nc : eq ] ) and ( [ second_ncgamma : eq ] ) , the sign of the determinant of the hessian matrix at the extremum is set by the sign of @xmath254 ; using eqs .
( [ dfdg : eq ] ) , ( [ dfzero : eq ] ) , and ( [ gamma : eq ] ) we obtain @xmath255 which is positive for @xmath256 but negative for @xmath257
. only the first solution corresponds to a minimum and thus to freezing , whereas the second is a saddle point . within the limits of the first - rlv - shell approximation ,
the crystals formed by @xmath1-potentials feature thus a _
universal localization parameter _ at freezing : irrespective of the location on the freezing line and even _ of the interaction potential itself _ , the localization length @xmath258 at freezing is a fixed fraction of the lattice constant and the parameter @xmath259 attains along the entire crystallization line the value @xmath260 we can understand the physics behind the constancy of the ratio @xmath261 by examining anew the variational form of the free energy , eq .
( [ fren_var : eq ] ) .
suppose we have a fixed density @xmath233 and we vary @xmath174 , seeking to achieve a minimum of @xmath196 .
an increase in @xmath174 implies an increase in the lattice constant @xmath193 by virtue of eq .
( [ den_a : eq ] ) .
the density profile takes advantage of the additional space created between neighboring sites and becomes more delocalized .
this increase of the spreading of the profile brings with it an entropic gain which exactly compensates the corresponding loss from the accumulation of particles on a single site , expressed by the term @xmath219 in eq .
( [ fren_var : eq ] ) . expressing @xmath258 in units of @xmath193 ,
i.e. , working with the variable @xmath215 instead with the original one , @xmath262 , brings the additional advantage that @xmath263 becomes independent of the pair potential .
the corresponding value of @xmath262 at freezing , @xmath264 , can be obtained from eqs .
( [ den_a : eq ] ) and ( [ proport : eq ] ) , and reads for the bcc - lattice as @xmath265 here , a dependence on the pair interaction appears through the value of @xmath230 .
complementary to the localization parameter , we can consider the lindemann ratio @xmath266 at freezing,@xcite taking into account that for the bcc lattice the nearest neighbor distance is @xmath267 .
employing the gaussian density parametrization , we find @xmath268 and thus @xmath269 . using ( [ gammaf : eq ] ) , the lindemann ratio at freezing , @xmath270 , is determined as @xmath271 this value is considerably larger than the typical value of @xmath272 usually quoted for systems with harshly repulsive particles , such as , e.g. , the bcc alcali metals and the fcc metals al , cu , ag , and au [ ] , but close to the value 0.160 found by stillinger and weber@xcite for the gaussian core model .
the particles in the cluster crystal are quite more delocalized than the ones for singly - occupied solids .
the clustering strategy enhances the stability of the crystal with respect to oscillations about the equilibrium lattice positions . the locus of freezing points @xmath273 is easily obtained by eqs .
( [ dfzero : eq ] ) and ( [ gamma : eq ] ) and takes the form of of a straight line : @xmath274 contrary to the lindemann ratio , which is independent of the pair potential , the freezing line does depend on the interaction potential between the particles . yet
, this dependence is a particular one , as it rests exclusively on the absolute value of the fourier transform at the minimum , @xmath275 and is simply proportional to it . comparing with the location of the @xmath124-line from eq .
( [ lambda : eq ] ) , @xmath276 , we find that crystallization _
preempts _ the occurrence of the instability : indeed , at fixed @xmath277 , @xmath278 or , equivalently , at fixed @xmath233 , @xmath279 ; see also fig .
[ phdg : fig ] .
the transition is first - order , as witnessed by the jumps of the values of @xmath176 and @xmath183 at the transition , which are nonzero for the crystal but vanish in the fluid .
this is analogous to the conjectured preemption of the kirkwood instability for infinite - dimensional hard spheres by a first - order freezing transition.@xcite the freezing properties of @xmath1-potentials are , thus , quite unusual and at the same time quite simple : the lattice constant is fixed due to a clustering mechanism that drives the aggregation number @xmath174 proportional to the density .
the constant of proportionality depends solely on the wavenumber @xmath230 for which the fourier transform of the pair interaction has a negative minimum , eq .
( [ proport : eq ] ) .
the freezing line is a straight line whose slope depends only on the value of the fourier transform of the potential at the minimum , eq .
( [ freeze : eq ] ) .
the lindemann ratio at freezing is a universal number , independent of interaction potential and thermodynamic state .
whereas the lindemann ratio is employed as a measure of the propensity of a crystal to melt , the height of the peak of the structure factor of the fluid is looked upon as a measure of the tendency of the fluid to crystallize .
the hansen - verlet criterion@xcite states that crystallization takes place when this quantity exceeds the value 2.85 . for the systems at hand ,
the maximum of @xmath280 lies at @xmath230 , as is clear from eq .
( [ sofk : eq ] ) . using eq .
( [ freeze : eq ] ) for the location of the freezing line , we obtain the value @xmath281 on the freezing line as @xmath282^{-1 } \cong 3.542.\ ] ] this value is considerably larger than the hansen - verlet threshold.@xcite in the fluid phase , @xmath1-systems can therefore sustain a higher degree of spatial correlation before they crystallize than particles with diverging interactions do .
this property lies in the fact that some contribution to the peak height comes from correlations from _ within _ the clusters that form in the fluid ; the formation of clusters already in the uniform phase is witnessed by the maxima of @xmath55 at @xmath283 seen in fig . [ gofr : fig ] and also explicitly visualized in our previous simulations of the model.@xcite these , however , do not contribute to intercluster ordering that leads to crystallization . at any rate ,
the hansen - verlet peak height is also a universal quantity for all @xmath1 systems , in the framework of the current approximation .
moreover , both for the lindemann and for the hansen - verlet criteria , the @xmath1 systems are more robust than usual ones , since they allow for stable fluids with peak heights that exceed @xmath284 by 25% and for stable crystals with lindemann ratios that exceed @xmath272 by almost 90% .
the density functional of eq .
( [ totalfren : eq ] ) is very accurate for the bounded ultrasoft potentials at hand.@xcite the modeling of the inhomogeneous density as a sum of gaussians is an approximation but , again , an accurate one , as has been shown by comparing with simulation results,@xcite see also section [ harmonic : sec ] of this work .
the analytical results derived in the preceding section rest on one additional approximation , namely on ignoring the rlvs beyond the first shell . here
, we want to compare with a full minimization of the functional ( [ totalfren : eq ] ) under the modeling of the density via ( [ gaussian_real : eq ] ) , so as to test the accuracy of the hitherto drawn conclusions on clustering and crystallization . ) , under the gaussian parametrization of the density , eq .
( [ gaussian_real : eq ] ) , redrawn from ref .
[ ] . on the same plot , we show by the dotted line the approximate analytical result for the freezing line , eq .
( [ freeze : eq ] ) , as well as the @xmath124-line of the system , eq .
( [ lambda : eq]).,width=321 ] we work with the concrete gem-4 system , for which the minimization of the density functional has been carried out and the phase diagram has been calculated in ref . [ ] . in fig .
[ phdg : fig ] we show the phase diagram obtained by the full minimization , compared with the freezing line from the analytical approximation , eq . ( [ freeze : eq ] ) , for this system .
it can be seen that the latter is a very good approximation to the full result , its quality improving slowly as the temperature grows ; the analytical approximation consistently overestimates the region of stability of the crystal .
moreover , whereas the approximation only predicts a stable bcc crystal , the high - density phase of the system is fcc .
although bcc indeed is , above the triple temperature , the stable crystal immediately post - freezing , it is succeeded at higher densities by a fcc lattice , which our analytical theory fails to predict . of the gem-4 potential .
main plot : a zoom at the region of @xmath12 in which the first few nonvanishing rlv shells of the cluster - crystals of fig .
[ phdg : fig ] lie .
the arrows denote the positions of the shells and the numbers in square brackets the numbers of distinct rlvs within each shell .
these positions are the result of the full minimization of the density functional .
downwards pointing arrows pertain to the direct bcc lattice and upwards pointing arrows to the direct fcc one . in agreement with clustering predictions ,
the positions of the rlvs are density - independent.,width=321 ] all these discrepancies can be easily understood by looking at the effects of ignoring the higher rlv shells from the summation in the excess free energy , eq .
( [ fren_var : eq ] ) . consider first exclusively the bcc lattice . in fig .
[ phiofq : fig ] we show the locations of the bcc - rlvs , as obtained from the full minimization , by the downwards pointing arrows .
it can be seen that the first shell is indeed located very closely to @xmath230 , as the analytical solution predicts . however , the next two rlv shells do have contributions and , due to their location on the hump of @xmath12 , the latter is positive . by ignoring them in performing the analytical solution
, we are artificially lowering the free energy of the crystal , increasing thereby its domain of stability .
the occurrence of a fcc - lattice that beats the bcc at high densities is only slightly more complicated to understand . a first remark is that the parameter @xmath262 increases proportionally to @xmath285 , see the following section .
hence , the gaussian factors from rlvs beyond the first shell , @xmath286 $ ] , @xmath287 , gain weight in the sum as density grows .
the cutoff for the rlv - sum is now provided rather by the short - range nature of @xmath12 than by the exponential factors .
due to the increased importance of the contributions from the @xmath287-terms in the excess free energy sum , the relative location of higher rlvs becomes crucial and can tip the balance in favor of fcc , although the bcc - lattice has a _ higher _ number of rlvs in its first shell than the fcc . in fig .
[ phiofq : fig ] we see that this is precisely what happens : the second rlv shell of the fcc is located fairly close to the first . in the full minimization , both of them arrange their positions so as to lie close enough to @xmath230 .
now , a total of fourteen 1st- and 2nd - shell rlvs of the fcc can beat the twelve 1st - shell rlvs of the bcc and bring about a structural phase transformation from the latter to the former . the relative importance of the first and second neighbors is quantified by the ratio @xmath288 , \label{ratio : eq}\ ] ] where @xmath289 is the number of rlvs in the second shell . if @xmath290 does not decay sufficiently fast to zero as @xmath291 grows , then the fcc lattice might even win over the bcc everywhere , since then @xmath292 could be considerable even for values of @xmath262 close to freezing , which are not terribly high .
in fact , the penetrable sphere model ( gem-@xmath17 with @xmath293 ) does not possess , on these grounds , a stable bcc phase at all.@xcite the prediction of the analytical theory on bcc stability has to be taken with care and is conditional to @xmath292 being sufficiently small . a quantitative criterion on the smallness of @xmath292 is model specific and can not be given in general .
the determination of the stable phases of the gem-@xmath17 family and their dependence on @xmath17 can be achieved by employing genetic algorithms@xcite and will be presented elsewhere.@xcite notwithstanding the quantitative discrepancies between the simplified , analytically tractable version of dft and the full one , which are small in the first place , the central conclusion of the former remains intact : the rlvs of the crystals are density - independent . whereas the analytical approximation predicts that the length of the first rlv shell coincides with @xmath230 , the numerical minimization brings about small deviations from this prediction .
however , by reading off the relevant values from fig .
[ phiofq : fig ] , we obtain @xmath294 for the bcc and @xmath295 for the fcc - lattice of the gem-4 model . comparing with the ideal value @xmath296 , we find that the deviation between them is only a few percent .
clustering takes place , so that the lattice constants of both lattices remain fixed , a characteristic that was also explicitly confirmed by computer simulations of the model.@xcite
the use of a gaussian parametrization for the one - particle density profiles , eq .
( [ gaussian_real : eq ] ) , is a standard modeling of the latter for periodic crystals .
this functional form is closely related to the harmonic theory of crystals.@xcite each particle performs oscillations around its lattice site , experiencing thereby an effective , one - particle site potential , @xmath297 that is quadratic in the displacement @xmath298 , for small values @xmath299 [ ] . here
, we will explicitly demonstrate that the gaussian form with the localization parameter predicted from density functional theory coincides with the results obtained by performing a harmonic expansion of the said site potential .
the formation of clustered crystals is a generic property of all @xmath1-systems , since the @xmath124-instability is common to all of them ; the form of the clusters that occupy the lattice sites , however , can be quite complex , depending on the details of the interaction .
the gaussian parametrization ( [ gaussian_real : eq ] ) implies that for each of the @xmath174 particles of the cluster , the lattice site @xmath300 is an equilibrium position . in other words ,
the particular clusters we consider here are internally structureless .
clusters with a well - defined internal order have been found when an additional hard core of small extent is introduced.@xcite a necessary requirement for the lack of internal order is that the laplacian of the interaction potential @xmath3 be finite at @xmath283 , as will be shown shortly . on these grounds
, we impose from the outset on the interaction potential the additional requirement : @xmath301 where the primes denote the derivative with respect to @xmath8 .
( [ condition1:eq ] ) implies that @xmath302 must be _ at least _ linear in @xmath8 as @xmath89 .
concomitantly , @xmath303 must be at least @xmath304 as @xmath89 . as a consequence
, we have @xmath305 it can be easily checked that ( [ condition : eq ] ) is satisfied by all members of the gem-@xmath17 class for @xmath20 .
it is also satisfied by the @xmath118 member , i.e. , the gaussian model , which does _ not _ display clustering because it belongs to the @xmath2-class .
this is , however , no contradiction . as mentioned above , the condition ( [ condition : eq ] ) is _ necessary _ for the formation of structureless clusters and not a sufficient one . for @xmath1-potentials for which ( [ condition : eq ] )
is not fulfilled ( such as the fermi distribution models of ref . [ ] ) , this does not mean that clusters do not form ; it rather points to the fact that they possess some degree of internal order . the clustered crystals can be considered as bravais lattices with a @xmath174-point basis . accordingly , their phonon spectrum will feature 3 acoustic modes and @xmath306 optical modes , for which the oscillation frequency @xmath307 remains finite as @xmath308 .
we are interested in the case @xmath182 , i.e. , deep in the region of stability of the crystal , where the clusters have a very high occupation number .
consequently , the phonon spectra and the particle displacements will be dominated by the optical branches .
further , we simplify the problem by choosing , in the spirit of the einstein model of the crystal,@xcite one specific optical phonon with @xmath309 as a representative for the whole spectrum .
this mode corresponds to the relative partial displacement of two sublattices : one with @xmath310 particles on each site and one with just the remaining one particle per site .
the two sublattices coincide at the equilibrium position and maintain their shape throughout the oscillation mode , consistent with the fact of an infinite - wavelength mode , @xmath311 .
accordingly , the site potential felt by any one of the particles of the single - occupied sublattice , @xmath297 , can be expressed as @xmath312 , \label{vsite : eq}\ ] ] where @xmath313 is the relative displacement of the two sublattices .
for brevity , we also define @xmath314 the taylor expansion of a scalar function @xmath315 around a reference point @xmath316 reads as @xmath317 setting @xmath318 and @xmath319 , we obtain the quadratic expansion of the site potential ; the constant @xmath320 is unimportant . for the linear term
, we have @xmath321 where @xmath322 and @xmath323 are unit vectors . the sum in ( [ linear : eq ] )
vanishes due to lattice inversion symmetry ; the first term also , since @xmath324 .
thus , the term linear in @xmath313 in the expansion of @xmath297 vanishes , consistently with the fact that @xmath325 is an equilibrium position .
we now introduce cartesian coordinates and write @xmath326 , @xmath327 , and @xmath328 .
the quadratic term in ( [ taylor3:eq ] ) takes the explicit form @xmath329 let us consider first the mixed derivative acting on @xmath330 , evaluated at @xmath331 . using eqs .
( [ wsite : eq ] ) and ( [ vsite : eq ] ) , we obtain @xmath332_{{\bf r } = 0 } \\ & + & \sum_{{\bf r}\ne 0}\frac{r_xr_y}{r^2 } \left[v''(r)-\frac{v'(r)}{r}\right ] .
\label{mixed : eq}\end{aligned}\ ] ] the first term on the right - hand side of ( [ mixed : eq ] ) vanishes by virtue of ( [ condition : eq ] ) .
the second one also vanishes due to the cubic symmetry of the lattice .
clearly , the other two terms in ( [ explicit : eq ] ) with mixed derivatives vanish as well . for the remaining terms ,
the cubic symmetry of the lattice implies that all three second partial derivatives of @xmath333 at @xmath331 are equal : @xmath334 gathering the results , we obtain the expansion of @xmath297 to quadratic order in @xmath298 as @xmath335s^2 , \label{isotropic : eq}\ ] ] which is isotropic in @xmath298 , as should for a crystal of cubic symmetry .
the one - particle motion is therefore harmonic ; as we consider @xmath182 , we set @xmath336 in ( [ isotropic : eq ] ) and we obtain the effective , one - particle hamiltonian @xmath337 in the form : @xmath338 with @xmath339 and the momentum @xmath340 and mass @xmath17 of the particle .
the density profile @xmath341 of this single - particle problem is easily calculated as @xmath342 , yielding @xmath343 this is indeed a gaussian of a single particle , with a localization parameter @xmath344 ; the total density on a given site will be then just @xmath345 , in agreement with the functional form put forward in eq .
( [ gaussian_real : eq ] ) .
it is useful to consider in detail the form of the localization parameter @xmath346 predicted by the harmonic theory .
the parameter @xmath347 is expressed as a sum of the values of @xmath348 over the periodic set @xmath173 . for every function
@xmath349 that possesses a fourier transform @xmath350 , it holds@xcite @xmath351 where @xmath185 is the set of rlvs of @xmath173 and @xmath352 is the density of _ lattice sites _ of @xmath173 . from ( [ nc : eq ] ) , @xmath353 .
taking into account that the fourier transform of @xmath354 is @xmath355 , we obtain for the localization parameter of the harmonic theory the result @xmath356 the localization parameter must be , evidently , positive .
( [ alphah : eq ] ) manifests the impossibility for cluster formation if the fourier transform of the pair potential is nonnegative , i.e. , for @xmath2 interactions . in the preceding section
, we showed within the dft formalism that if the potential is @xmath1 , this implies the formation of cluster crystals .
harmonic theory allows us to make the opposite statement as well : if the potential is _ not _
@xmath1 , then there can be _ no clustered crystals_. therefore , an _ equivalence _ between the @xmath1 character of the interaction and the formation of clustered crystals can be established .
moreover , eq . ( [ alphah : eq ] ) offers an additional indication as to why the rlv where @xmath357 is most negative is selected as the shortest nonvanishing one by the clustered crystals : this is the best strategy in order to keep the localization parameter positive .
harmonic theory provides , therefore , an insight into the necessity of locating the first shell of the rlvs at @xmath123 from a different point of view than density functional theory does .
the choice @xmath358 guarantees that the particles inhabiting neighboring clusters provide the restoring forces that push any given particle back towards its equilibrium position .
the density functional treatment of the preceding sections establishes that the lattice constant is chosen by @xmath1-systems in such a way that the sum of _ intracluster _ and _ intercluster _ interactions , together with the entropic penalty for the aggregation of @xmath174 particles is optimized.@xcite the unlimited growth of @xmath174 is avoided by the requirement of mechanical stability of the crystal . indeed , for too high @xmath174-values , the lattice constant would concomitantly grow ,
so that the resulting restoring forces working against the thermal fluctuations , would become too weak to sustain the particles at their equilibrium positions .
let us , finally , compare the result ( [ alphah : eq ] ) for the localization parameter with the prediction from dft .
we consider the high - density crystal phase , for which @xmath262 is very large , so that the simplification that only the first shell of rlvs can be kept in ( [ fren_var : eq ] ) must be dropped .
setting @xmath359 there , we obtain @xmath360 the function @xmath12 is short - ranged in reciprocal space , thus the sum in ( [ dftall : eq ] ) can be effectively truncated at some finite upper cutoff @xmath361 . then , there exists a sufficiently large density @xmath233 beyond which the parameter @xmath262 is so large that @xmath362 for all @xmath363 included in the summation .
accordingly , we can approximate all exponential factors with uniti in ( [ dftall : eq ] ) , obtaining an algebraic equation for @xmath262 . reverting back to dimensional quantities
, its solution reads as @xmath364 and is _ identical _ with the result from the harmonic theory , eq .
( [ alphah : eq ] ) .
thus , density functional theory and harmonic theory become identical to each other at the limit of high localization .
this finding completes and generalizes the result of archer,@xcite who established a close relationship between the mean - field dft and the einstein model for the form of the variational free energy functional of the system .
consistently with our assumptions , @xmath176 indeed grows with density .
in fact , since the set of rlvs in the sum of ( [ alphasdft : eq ] ) is fixed , it can be seen that @xmath176 is simply proportional to @xmath365 .
this peculiar feature of the class of systems we consider is not limited to the localization parameter , and its significance is discussed in the following section .
as can be easily confirmed by the form of the density functional of eq .
( [ totalfren : eq ] ) , the mean - field nature of the class of ultrasoft systems considered here ( both @xmath2- and @xmath1-potentials ) implies that the structure and thermodynamics of the systems is fully determined by the ratio @xmath366 between density and temperature and not separately by @xmath40 and @xmath367 .
this is a particular type of scaling between the two relevant thermodynamic variables , reminiscent of the situation for systems interacting by means of inverse - power - law potentials @xmath3 having the form : @xmath368 for such systems , it can be shown that their statistical mechanics is governed by a a single coupling constant @xmath369 expressed as@xcite @xmath370 where @xmath371 is the space dimension .
it would appear that inverse - powers @xmath372 satisfy precisely the same scaling as mean - field systems do but there is a condition to be fulfilled : inverse - power systems are stable against explosion , _ provided _ that @xmath373 ; this can be most easily seen by considering the expression for the excess internal energy per particle , @xmath374 , given by@xcite @xmath375 ! } \int_0^{\infty}r^{d-1-n}g(r;\rho , t){\rm d}r .
\label{inten : eq}\ ] ] as @xmath376 for @xmath377 , we see that the integral in ( [ inten : eq ] ) converges only if @xmath373 ; a logarithmic divergence results for @xmath372 .
a ` uniform neutralizing background ' has to be formally introduced for @xmath378 , to obtain stable pseudo one - component systems , such as the one - component plasma.@xcite since we aim at staying with genuine one - component systems throughout , we must strictly maintain @xmath373 . instead of taking the limit @xmath379 , we consider therefore a different procedure by setting @xmath380 with some arbitrary , finite @xmath381 .
then the coupling constant @xmath369 becomes @xmath382 now take @xmath163 in this prescribed fashion , obtaining : @xmath383 which has precisely the same form as the coupling constant of our systems . in taking the limit @xmath163 , the exponent @xmath384 of the inverse - power potential diverges as well .
it can be easily seen that in this case , the interaction @xmath3 of eq .
( [ powerlaw : eq ] ) becomes a hard - sphere potential of diameter @xmath6 .
in other words , the procedure prescribed above brings us once more to infinite - dimensional hard spheres , because also the inverse - power interaction becomes arbitrarily steep .
this is a very different way of taking the limit than in refs .
[ ] : there , the interaction is hard in the first place and subsequently the limit @xmath163 is taken , whereas here , interaction and dimension of space change together , in a well - prescribed fashion .
the fact that the statistical mechanics of ultrasoft fluids in three dimensions is determined by the same dimensionless parameter as that of a particular realization of hard spheres in infinite dimensions is intriguing . in a sense ,
ultrasoft systems are effectively high - dimensional , since they allow for extremely high densities , for which every particle interacts with an exceedingly high number of neighbors .
they might , in this sense , provide for three - dimensional approximate realizations of infinite - dimensional models .
this is yet another relation to infinite - dimensional systems , in addition to the one discussed at the end of sec .
[ dft : sec ] . whether there exists a deeper mathematical connection between the two classes , remains a problem for the future .
we have provided a detailed analysis of the properties of bounded , ultrasoft systems , with emphasis on the @xmath1-class of interaction potentials .
after having demonstrated the suppression of the contributions from the high - order direct correlation functions of the fluid phases ( of order 3 and higher ) , we established as a consequence the accurate mean - field density functional for arbitrary inhomogeneous phases .
though this functional has been introduced and successfully used in the recent past both in statics@xcite and in dynamics,@xcite a sound justification of its basis on the properties of the uniform phase was still lacking . the persistence of a _ single _ , finite length scale for the lattice constants of the ensuing solids of @xmath1-systems has been understood by a detailed analysis of the structure of the free energy functional . in the fluid ,
the same length scale appears since the position of maximum of the liquid structure factor is independent of density . the negative minimum of the interaction potential in fourier space sets this unique scale and forces in the crystal the formation of clusters , whose population scales proportionally with density .
the analytical solution of an approximation of the density functional is checked to be accurate when confronted with the full numerical minimization of the latter .
universal lindemann ratios and hansen - verlet values at crystallization are predicted to hold for all these systems , which differ substantially from those for hard matter systems .
the analytical derivation of these results provides useful insight into the robustness of these structural values for an enormous variety of interactions . though the assumption of bounded interactions has been made throughout , recent results@xcite
indicate that both cluster formation and the persistence of the length scale survive when a short - range diverging core is superimposed on the ultrasoft potential , provided the range of the hard core does not exceed , roughly , 20% of the overall interaction range.@xcite the morphology of the resulting clusters is more complex , as full overlaps are explicitly forbidden ; even the macroscopic phases are affected , with crystals , lamellae , inverted lamellae and ` inverted crystals ' showing up at increasing densities . the generalization of our density functional theory to such situations and the modeling of the nontrivial , internal cluster morphology is a challenge for the future . here , a mixed density functional , employing a hard - sphere and a mean - field part of the direct correlation function seems to be a promising way to proceed.@xcite finally , the study of the vitrification , dynamical arrest and hopping processes in concentrated @xmath1-systems is another problem of current interest .
the recent ` computer synthesis ' of model , amphiphilic dendrimers that do display precisely the form of @xmath1-interactions discussed in this work,@xcite offers concrete suggestions for the experimental realization of the hitherto theoretically predicted phenomena .
we are grateful to andrew archer for helpful discussions and a critical reading of the manuscript and to andras st for helpful comments .
this work has been supported by the sterreichische forschungsfond under project no .
p17823-n08 , as well as by the deutsche forschungsgemeinschaft within the collaborative research center sfb - tr6 , `` physics of colloidal dispersions in external fields '' , project section c3 .
consider the inverse fourier transform of the spherically symmetric , bounded pair potential @xmath3 , reading as @xmath385 from ( [ invft : eq ] ) , it is straightforward to show that the second derivative of @xmath3 at @xmath283 takes the form @xmath386 evidently , if @xmath387 , then @xmath15 _ must _ have negative parts and hence @xmath3 is @xmath1 . for the gem-@xmath17 family ,
it is easy to show that @xmath388 for @xmath20 , thus these members are indeed @xmath1 , as stated in the main text .
double - gaussian potentials of the form @xmath389 with @xmath390 , @xmath391 , which feature a local _ minimum _ at @xmath56 are also @xmath1 , for the same reason .
notice , however , that @xmath387 is a sufficient , not a necessary condition for membership in the @xmath1-class .
thus , there exist @xmath1-potentials for which @xmath392 .
the introduction of the new variable @xmath215 instead of @xmath262 , eq .
( [ amin_nc : eq ] ) , and the subsequent new form @xmath223 of the variational free energy , eq .
( [ fbar : eq ] ) , are just a matter of convenience , which makes the minimization procedure more transparent .
a free gift of the variable transformation is also the ensuing diagonal form of the hessian matrix at the extremum .
fully equivalent results are obtained , of course , by working with the original variational free energy , @xmath196 , eq .
( [ fren_var : eq ] ) . here
we explicitly demonstrate this equivalence .
keeping , consistently , only the first shell of rlvs with length @xmath220 , @xmath196 takes the form : @xmath393 \\ \nonumber & + &
\frac{\rho^*}{2}\tilde\phi(0 ) \\ & + & \frac{\xi_1\rho^*}{2}\tilde\phi(y_1 ) e^{-y_1 ^ 2/(2\alpha^ * ) } , \label{fren_new : eq}\end{aligned}\ ] ] where @xmath220 and @xmath174 are related via eq . (
[ d : eq ] ) .
minimizations of @xmath196 with respect to @xmath174 and @xmath262 yield , respectively : @xmath394e^{-y_1 ^ 2/(2\alpha^ * ) } = 0 , \label{delnc : eq}\ ] ] and @xmath395 subtracting the last two equations from one another we obtain @xmath396 implying @xmath397 , as in the main text ( once more , @xmath226 is a formal solution that must be rejected on the same grounds mentioned in the text . ) from this property , eq .
( [ proport : eq ] ) immediately follows . introducing @xmath398
, we can determine @xmath264 on the freezing line by requiring the simultaneous satisfaction of the minimization conditions above and of @xmath399 .
the latter equation yields @xmath400\right ] \\
= - \frac{\xi_1\rho^*}{2}\tilde\phi(y_*)e^{-y_*^2/(2\alpha^*_{\rm f } ) } , \label{delftildezero : eq}\end{aligned}\ ] ] which , together with eq .
( [ dela : eq ] ) , yields , after some algebra @xmath401 - 1 = \frac{2\alpha^*_{\rm f}}{y_*^2}. \label{onemore : eq}\ ] ] using eqs .
( [ amin_nc : eq ] ) and ( [ y1nc : eq ] ) , we obtain the equation for the @xmath263-parameter at freezing as @xmath402 = \frac{2}{\gamma_{\rm f}\zeta^2 } , \label{gamma_new : eq}\ ] ] which , upon setting @xmath241 and @xmath242 , yields eq .
( [ gamma : eq ] ) of the main text .
alternatively , we can introduce the variable @xmath403 and rewrite eq .
( [ onemore : eq ] ) as @xmath404 - 1 = 2 t , \label{omega : eq}\ ] ] which delivers @xmath405 as a solution or , equivalently , @xmath406 , in agreement with eqs .
( [ gammaf : eq ] ) and ( [ alphaf : eq ] ) of the main text . strictly speaking
, there is no freezing _ line _ on the density - temperature plane but rather a domain of coexistence , since it is equality of the grand potentials that determines phase coexistence .
nevertheless , the condition of equality of the helmholtz free energies determines a line that runs between the melting- and crystallization lines and therefore serves as a reliable locator of the freezing transition .
a very important exception is the hard - sphere potential , for which no harmonic expansion can be made .
the fact that hard - sphere crystals do still have density profiles that are very well modeled by gaussians , calls for a different explanation there .
it can be argued that the huge number of uncorrelated collisions with the neighbors , together with the central limit theorem , are responsible for the gaussianity of density profiles in hard - sphere solids . | we demonstrate the accuracy of the hypernetted chain closure and of the mean - field approximation for the calculation of the fluid - state properties of systems interacting by means of bounded and positive - definite pair potentials with oscillating fourier transforms .
subsequently , we prove the validity of a bilinear , random - phase density functional for arbitrary inhomogeneous phases of the same systems . on the basis of this functional , we calculate analytically the freezing parameters of the latter .
we demonstrate explicitly that the stable crystals feature a lattice constant that is independent of density and whose value is dictated by the position of the negative minimum of the fourier transform of the pair potential .
this property is equivalent with the existence of clusters , whose population scales proportionally to the density .
we establish that regardless of the form of the interaction potential and of the location on the freezing line , all cluster crystals have a universal lindemann ratio @xmath0 at freezing .
we further make an explicit link between the aforementioned density functional and the harmonic theory of crystals .
this allows us to establish an equivalence between the emergence of clusters and the existence of negative fourier components of the interaction potential .
finally , we make a connection between the class of models at hand and the system of infinite - dimensional hard spheres , when the limits of interaction steepness and space dimension are both taken to infinity in a particularly described fashion . |
facility location is one of the most studied optimization problems in operations research and captures a large variety of applications .
a classical motivation is placing facilities ( e.g , warehouses ) in such a way that the combined costs of customer satisfaction and warehouse construction are minimized . however , there are also plenty of applications in distributed scenarios . for instance , in wireless networks a set of nodes has to be chosen to provide some services ( e.g , a distributed database ) . making such services
available incurs costs at those _ facility _ nodes , while all remaining nodes act as _
clients_. they use the services of the nearest facility node , and are charged a cost proportional to the corresponding distance .
the objective is to determine a set of facility nodes such that the costs caused by the facilities and the clients is as low as possible .
[ [ formal - problem - definition . ] ] formal problem definition .
+ + + + + + + + + + + + + + + + + + + + + + + + + + we consider the _ metric uncapacitated facility location _ problem in a distributed setting . here , we are given a complete bipartite graph @xmath5 consisting of a set of _ facilities _ @xmath6 and a set of _ clients _ @xmath7 . to each facility
@xmath8 an opening cost @xmath9 is assigned .
each edge @xmath10 in @xmath11 is weighted with the value @xmath12 that represents the costs of connecting client @xmath13 with facility @xmath14 .
the objective is to determine a subset of the facilities to be opened and connect every client to at least one open facility in such a way that the sum of the opening costs and connection costs is minimized .
the linear program representation of the facility location problem and its dual program are as follows : @xmath15 @xmath16 the variable @xmath17 indicates whether facility @xmath14 is open ( @xmath18 ) or closed ( @xmath19 ) .
the other indicator variable @xmath20 has the value @xmath21 if the client @xmath13 is connected to facility @xmath14 , and @xmath22 otherwise . the constraints ( @xmath21 ) guarantee that each client is connected to at least one facility , while the constraints ( @xmath23 ) make sure that a client can only be connected to an open facility .
the problem we consider is _ metric _ , since the values @xmath24 are required to satisfy the triangle inequality ( i.e , @xmath25 ) , and _ uncapacitated _ , since an arbitrary number of clients can be connected to an open facility .
furthermore , for the sake of presentation we assume all @xmath26 to be normalized such that the smallest non - zero value is @xmath21 .
the dual program will be used in the description of our approximation algorithm .
intuitively , a @xmath27 variable can be seen as the amount the client @xmath13 is willing to pay for being connected to a facility . from the point of view of a facility @xmath14
, @xmath27 is the sum of @xmath24 ( the amount @xmath13 pays for a connection to @xmath14 ) and @xmath28 ( the amount it pays for opening @xmath14 ) .
our algorithm is executed in the @xmath0 model , which was introduced by peleg @xcite and is commonly used to model the execution of distributed algorithms on graphs : algorithms are executed in synchronous send - receive - compute rounds . in a single round , each node sends a message to each of its neighbors in the graph . note that the messages sent to each neighbor by a single node are not required to contain the same information .
once all nodes have sent their messages , they receive a single message from each of their neighbors .
after all the messages have been received , every node is allowed to spend an arbitrary amount of time for computation ( i.e , computation is for free and we are only interested in the number of communication rounds ) .
the end of the computation by all nodes marks the start of a new send - receive - compute round .
the message size in the @xmath0 model is bounded .
we will limit the size of the messages used in our algorithm to @xmath29 bits .
the same limitations were also used in @xcite , @xcite and @xcite .
this bound is reasonable , because it allows the nodes to send their i d in a single message .
due to this constraint on message size , we restrict the values @xmath30 and @xmath24 in such a way that it is possible to represent them with @xmath29 bits ( i.e. to be able to send them in a single message ) .
the graph our algorithm is executed on is the complete bipartite graph of clients and facilities .
this means that within a single round each client can communicate with all facilities and each facility can communicate with all clients . since nodes of the same partition can not communicate directly with each other , but instead have to use nodes of the other partition as relays , gathering all information about the graph requires @xmath31 rounds
note that without a limit on the message size our problem could be trivially solved in four rounds ( all information about the graph is gathered at a single node in two rounds , an optimal solution is computed and distributed in two more rounds ) .
we want to stress that , although in our algorithm the nodes communicate with all their neighbors in each round , it is possible to restate the algorithm such that nodes @xmath14 and @xmath13 where @xmath24 is `` large '' ( more than @xmath32 ) never communicate with each other .
[ [ our - contribution . ] ] our contribution .
+ + + + + + + + + + + + + + + + + we present a distributed approximation algorithm for the metric uncapacitated facility location problem .
it is based on a greedy algorithm by jain et al@xcite ( from here on referred to as ) and yields a guaranteed ( not in expectation ) @xmath33-approximation in @xmath3 rounds ( with high probability ) .
our algorithm is executed in the @xmath0 model on a complete bipartite graph .
although it is strongly related to the _ greedyfl _ algorithm , there are ( due to the parallel execution ) new challenges regarding the selection process that occurs when multiple facilities are eligible for opening .
the difficulty is that we require the cost of the solution computed by our algorithm to be always at most @xmath33 times worse than an optimal solution ( in contrast to similar work by blelloch et al@xcite , where the approximation factor is increased by factor @xmath23 ) .
moreover , unlike other work ( e.g , @xcite , where the approximation factor is given `` in expectation '' ) , our algorithm provides a worst - case guarantee on the approximation factor .
[ [ related - work . ] ] related work .
+ + + + + + + + + + + + + during the last two decades , the uncapacitated metric facility location problem was of great interest , and a lot of progress has been made concerning the running time and approximation factor of sequential algorithms solving it .
aardal et alintroduced the first polynomial time algorithm yielding a @xmath34-approximation @xcite .
improving this approximation factor was the topic of a multitude of research papers .
for example , chudak et alimproved the approximation factor to @xmath35 @xcite .
the factor has been improved , until eventually byrka designed a @xmath36-approximation algorithm @xcite , which , at this point in time , yields the best known approximation .
although both last mentioned algorithms yield very good approximation factors , they have , due to applying lp - rounding , high running times .
thus , the design of algorithms with slightly worse approximation factors , but better running times , was also of interest in the past .
jain and vazirani used the primal - dual approach to develop and analyze an algorithm with approximation factor @xmath37 and a running time of @xmath38 @xcite .
a simplified and faster @xmath39 version of this algorithm was introduced by mettu and plaxton @xcite .
later on , jain et alimproved the former results by presenting two algorithms in @xcite : one with running time @xmath40 and an approximation factor of @xmath41 ( which we refer to as ) and another one with running time @xmath42 and an approximation factor of @xmath43 ( @xmath4 denoting the number of nodes and @xmath44 the number of edges in the complete bipartite graph of facilities and clients ) .
later on , building upon @xcite , mahdian et alimproved the factor to @xmath45 @xcite . under the assumption that @xmath46 , guha et alshowed in @xcite that no polynomial time algorithm with an approximation factor better than @xmath47 exists .
thus , an improvement of either the lower or the upper bound for the approximation factor , even in the @xmath37rd or higher fractional digit , is of great interest .
the following results concerning the facility location problem can be found in the distributed scenario : in @xcite pandit et alpresent an algorithm yielding a @xmath48-approximation and a running time of @xmath29 .
their algorithm is a parallel version of the primal - dual algorithm by jain et al@xcite and is like our algorithm executed in the @xmath0 model on a complete bipartite graph with message size limited to @xmath29 bits .
pandit et alassume that the difference between the opening costs of the cheapest and the most expensive facility can be arbitrary large .
it is reasonable to drop this assumption , since they ( as well as we ) require that the facility opening costs and distances between clients and facilities are encoded with @xmath29 bits . by dropping this assumption and modifying the algorithm and analysis by pandit et alin a small way ( changing a factor from @xmath23 to @xmath49 ) , it is possible to achieve a @xmath50 approximation factor in @xmath51 rounds .
further improvement of the factor with this approach is not possible , since the approximation factor of @xcite , which pandit et alparallelized , is @xmath37 .
this means that their algorithm , while faster than ours , has a worse approximation factor .
recently , in @xcite pandit et alall presented a technique that can be used to execute greedy facility location algorithms ( like ) in parallel in polylogarithmic time .
although they consider a similar problem as we do , their results are quite different from ours : their approximation factor is @xmath52 in expectation ( their algorithm can produce an arbitrary bad solution , even though this is very unlikely ) , while we can guarantee an approximation factor of @xmath33 in the worst case . also ,
although they do not state their exact approximation factor , the expected approximation factor achieved with their technique can not , to the best of our knowledge , be decreased below @xmath53 , where @xmath54 is the approximation factor of the sequential algorithm ( i.e , @xmath55 if used with ) .
a similar result to @xcite was presented by blelloch et alin @xcite . instead of
the @xmath0 they use a pram model and achieve a @xmath56 approximation with running time @xmath57 by parallelizing the algorithm . in order to choose which facilities to open , they use a technique introduced by rajagopalan and vazirani @xcite .
this technique can also be used to execute greedy facility location algorithms ( in a distributed manner and polylogarithmic time ) and yields a guaranteed approximation factor of @xmath58 , where @xmath54 is again the approximation factor of the corresponding sequential algorithm ( i.e@xmath59 if used with ) .
other results that also use the bound of @xmath29 on the message complexity are @xcite and @xcite . in @xcite moscibroda
et alshow that in @xmath60 communication rounds a @xmath61 approximation in @xmath60 communication rounds ( @xmath62 and @xmath63 are the number of facilities , respectably clients , and @xmath64 a coefficient dependent on instance parameters ) can be achieved in the more general _ non - metric _ facility location problem .
gehweiler et alshow in @xcite that @xmath37 rounds are sufficient to compute a @xmath52-approximation for the uniform facility location problem ( opening each facility incurs the same costs ) .
@xcite applies the approach of mettu and plaxton @xcite , which has been successful in a lot of other settings as well : in the kinetic setting @xcite , in game theoretic settings @xcite , for algorithms working in sublinear time @xcite , and when confronted with perpetual changes to the problem instance @xcite .
[ [ structure - of - the - paper . ] ] structure of the paper .
+ + + + + + + + + + + + + + + + + + + + + + + in section 2 we give a detailed description of our approximation algorithm and prove its approximation factor by generalizing the techniques and results of @xcite .
our main contribution is section 3 . here
, we deal with the problem of selecting facilities to be opened that arises due to the parallel execution of the _ greedyfl _ algorithm by jain et aland show that our distributed algorithm presented in section 2 terminates in sublinear time .
our parallel algorithm is based on the _ greedyfl _ algorithm by jain et al@xcite .
as stated above , it is executed in the @xmath0 model on a complete bipartite graph .
we will first describe the algorithm and then give an analysis of its approximation factor and running time .
[ [ algorithm - description . ] ] algorithm description . + + + + + + + + + + + + + + + + + + + + + + due to the completeness of the bipartite graph , in a single synchronous round each client can send to and receive a message from all facilities .
the same applies analogously to each of the facilities . during the algorithm s execution
, clients and facilities can be in various states : clients can be _ not connected _ or _ connected _ , while facilities can be _ closed _ , _ currently - paid _ , or _
open_. intuitively , _
currently - paid _ facilities compete with each other for the permission to change their state to _
all clients start in the _ not connected _ state and all facilities in the _ closed _ state .
when a client @xmath13 changes its state to _ connected _ , it is assigned to an _ open _ facility @xmath14 ( @xmath65 ) and never changes its state again .
a _ closed _ facility can only change its state to _ currently - paid _
, while a _ currently - paid _ facility can either become _ closed _ or @xmath66 .
once a facility is @xmath66 , it never changes its state .
the algorithm terminates , once all clients are _
connected_. the solution returned by our algorithm is the set of all _ open _ facilities and the assignment @xmath67 .
each facility knows now whether it is open or not and each client knows the facility it is assigned to .
@xmath68 $ ] + execute algorithm 4 until there are no _ currently - paid _ facilities left compute @xmath69 the algorithm is executed in a loop and each passing of the loop is referred to as a _
phase_. the behavior of a client during a phase is described in algorithm 1 , while the behavior of a facility is described in algorithm 2 .
facilities and clients interact with each other by transmitting messages ; to each send operation there exists a corresponding receive operation .
the idea of the algorithm is to assign to each client @xmath13 a variable @xmath27 initialized with @xmath21 .
this variable represents the amount a client is willing to pay for being connected to a facility and for opening this facility in the current phase ( i.e , the @xmath27 variable of the dual program ) .
each _ not connected _
client @xmath13 increases this variable in every phase by multiplying it with @xmath49 where @xmath2 is an arbitrary small constant greater than @xmath22 ( fixed at the start of our algorithm ) .
after @xmath27 is increased , all clients send their current @xmath27 to all facilities .
@xmath70 $ ] @xmath71 compute @xmath72 @xmath73 _ currently - paid _ + execute algorithm 3 until there are no _ currently - paid _ facilities left let @xmath74 be the set of all _ not connected _ clients and @xmath75 .
we say that client @xmath13 _ contributes _ to facility @xmath14 , if @xmath76 ( the client pays @xmath24 for being connected to and @xmath28 for the opening of facility @xmath14 ) . upon receiving the @xmath27 variables ,
every _ closed _
facility @xmath14 computes @xmath77 , which represents the amount _ not connected _ clients are willing to contribute to the payment of opening facility @xmath14 .
if the contribution to the opening of facility @xmath14 ( i.e. @xmath77 ) reaches a point such that @xmath78 , it changes its status to _ currently - paid_. @xmath79 uniformly distributed random number in @xmath80 $ ] @xmath81 $ ] + @xmath82 $ ] define @xmath83 where @xmath84 + @xmath85 @xmath86 @xmath87 @xmath88 $ ] + @xmath89 $ ] + @xmath90 @xmath91 opening all the _ currently - paid _ facilities could result in a very bad approximation , since it is possible that a single _ not connected _
client @xmath13 is contributing to an arbitrary number of _ currently - paid _ facilities ( i.e. @xmath92 ) and opening just one of these facilities would be sufficient .
thus , a selection procedure is needed to determine which _ currently - paid _ facilities are going to be permanently opened ( change their status to _ open _ ) and which of them will remain closed in this phase ( changing their status from _ currently - paid _ to _ closed _ ) .
furthermore , it is required to determine which client is connected to which facility .
all of this is achieved by the interaction of algorithm 3 ( from the point of view of facility @xmath14 ) and algorithm 4 ( from the point of view of client @xmath13 ) . in each iteration of algorithm 3 and 4 ,
a subset of _ currently - paid _ facilities is chosen to be opened .
the development and analysis of these two algorithms is the main contribution of our paper , as they are responsible for selecting the facilities ( which are going to be opened ) in a fast and efficient way .
the algorithms operate by generating a random number at each facility and then opening a facility , if its number is the highest one in its neighborhood ( the neighborhood of a facility @xmath14 consists of all facilities @xmath93 , such that there is at least one client that contributes to @xmath14 and @xmath93 ) .
only @xmath29 bits are used to represent the generated random number , which is sufficient to guarantee that , with high probability , no two facilities generate the same number .
the main objective of algorithm 3 and 4 is to establish the following fact : in every phase , after the iterative execution of algorithm 3 and 4 terminates , the following is true for all facilities : 1 .
no facility is _ currently - paid_. 2 . if facility @xmath14 is _ open _ , then @xmath94 , where @xmath95 .
3 . if facility @xmath14 is _ closed _ , then @xmath96 , where @xmath97 . intuitively , ( ii ) means that a facility is only opened , if all clients connected to it can ( together ) pay for its opening , and ( iii ) means that , after a phase ended , there is no _ closed _ facility to which clients contribute enough to pay for its opening . the fact that a _ closed _ facility with @xmath98 can not enter the next phase , as implied by ( iii ) , is of importance .
it ensures that the desired approximation factor can be guaranteed , which is curcial for ( the later introduced ) lemma 2 .
@xmath99 $ ] + define @xmath100 where @xmath101 + @xmath102 $ ] + @xmath103 $ ] + @xmath104 + @xmath105 _ connected _
@xmath106 @xmath107 @xmath108 $ ] + note that after the algorithm s termination the cost of the solution computed by our algorithm lies in the interval @xmath109 $ ] .
this is due to the following facts : each client is connected to exactly one facility and this facility is _
open_. since @xmath110 a client @xmath13 contributes to the opening of a facility @xmath14 only after @xmath27 covered the costs @xmath24 of connecting @xmath14 to @xmath13 .
also , since we require that only after @xmath78 facility @xmath14 changes its state to _ open _ , the opening of @xmath14 is completely paid for and since @xmath74 is the set of all _ not connected _ clients , it is made sure that a client contributes to the opening of at most one facility .
the running time of the entire algorithm is bounded by @xmath3 rounds : the number of phases is @xmath51 , since in the worst case a client has distance @xmath111 to its nearest facility and this facility has opening costs of @xmath111 ( the @xmath24 and @xmath30 are limited by @xmath29 bits ) .
thus , increasing the @xmath112 value iteratively by multiplying it with @xmath49 , where @xmath113 , results in @xmath51 rounds .
later , we will show that with high probability algorithm 3 and 4 terminate after @xmath114 rounds . the value chosen for @xmath2 not only determines the runtime , but also the approximation factor , which is @xmath115 .
[ [ analysis - of - the - approximation - factor - and - runtime . ] ] analysis of the approximation factor and runtime .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + intuitively , the algorithm can be seen as continuously and simultaneously increasing the @xmath27 values .
this allows to open a facility @xmath14 in the exact moment it is paid for ( i.e. , when @xmath116 ) ; a property which is essential for proving its approximation factor .
in contrast , our algorithm increases the @xmath27 values in discrete steps and thus , in general , it may occur that @xmath98 .
the goal of this section is to show that even though our algorithm may yield a very different solution from the sequential algorithm
the costs of the optimal and our solution do not differ too much .
namely , we show that our approximation factor is at most @xmath117 times greater than the approximation factor of .
our proof uses similar arguments as , and is based on , the one of jain et al , but it turns out that in order to apply the results of jain et al , we need to get a deeper insight into the structural properties of the linear programs involved .
the @xmath27 and @xmath28 values computed by our algorithm can be interpreted as a solution to the dual program of the facility location program presented in the introduction .
however , they do not necessarily represent a feasible solution : our algorithm opens a facility when @xmath118 ( summing over all _ not connected _ clients ) and not , as required by constraint ( 4 ) , over all the clients . if we choose @xmath119 such that @xmath120 and @xmath121 constitute a feasible dual solution , the duality theorem yields that our solution is by a factor of at most @xmath119 more expensive than an optimal solution . for their original , sequential algorithm jain et algave an upper bound of @xmath41 for @xmath119 @xcite .
they denoted the used technique as dual fitting with a factor - revealing lp .
we will construct such a factor revealing program ( see ) for our algorithm and establish a structural connection between it and the lp by jain et al(see ) for their sequential algorithm ( ) . to this end
, consider a single facility @xmath14 and assume , that w.l.o.g@xmath122 holds only for the first @xmath123 clients and that @xmath124 .
the next two lemmas express the constraints imposed by both the algorithm and the facility location problem .
they resemble similar lemmas to those used in the proof for the sequential algorithm ( cf .
@xcite ) ( their proofs can be found in the appendix ) .
[ lem : metric_lp_property ] for any two clients @xmath13 , @xmath125 and a facility @xmath14 , we have @xmath126 .
[ lem : facilities_not_overpaied_property ] for every client @xmath13 and facility @xmath14 , the execution of the algorithm guarantees that @xmath127 .
our algorithm ensures that the constraints stated by the two lemmas above are met .
thus , we now want to find the minimum value for @xmath119 such that @xmath128 holds , without violating the lemmas statements .
equivalently , we want to maximize the ratio @xmath129 ( again , without violating the constraints ) .
this formulation yields the following family of linear programs , referred to as the factor - revealing lp .
@xmath130 @xmath131 note that constraints @xmath132 resp@xmath133 are only used to sort the @xmath27 values in order to ease the formulation of constraints @xmath134 resp@xmath135 .
the following theorem quantifies the correlation between our ( ) and the original factor - revealing lps ( ) . for any @xmath136 , the optimal solution to
is bounded by @xmath137 , and thus the approximation factor of our algorithm is also @xmath137 .
jain et alshowed that @xmath138 can be at most @xmath41 in . in our modified version @xmath138 can be at most @xmath137 : fix a problem instance by setting the @xmath24 variables and the @xmath30 variable to arbitrary values @xmath139 and consider the following two propositions .
[ propositionlp1 ] in an optimal solution @xmath140 to , @xmath141 at least one of the two properties is true : 1 .
@xmath142 , where @xmath143 2 .
@xmath144 and @xmath145 note that the term @xmath146 is fixed . since we deal with a maximization problem , which is bounded ( see constraint @xmath134 of ) , each @xmath27 is bounded .
since @xmath147 represents an optimal solution , for each @xmath27 there must be at least one constraint that bounds @xmath27 and is also tight .
if this tight constraint is of the form @xmath148 , ( i ) holds .
otherwise , there must be at least one tight constraint of the form @xmath149 , i.e. @xmath150 .
this argument can be analogously applied to @xmath151 : either @xmath152 , or there exists a @xmath153 such that @xmath154 . like this
, we can build a recursion for @xmath155 , which terminates as soon as we reach a variable @xmath156 with @xmath157 and thus ( i ) holds for @xmath158 .
note that for every @xmath27 there must be such a terminating sequence , since constraint @xmath134 is the only constraint which gives an absolute upper bound on the variables .
otherwise , the @xmath27 would be unbounded .
the recursion yields @xmath159 , implying @xmath160 .
since also @xmath161 ( constraint @xmath162 ) , we have @xmath163 , proving ( ii ) .
[ propositionlp2 ] given an optimal solution @xmath164 to and an optimal solution @xmath165 to , we have @xmath166 .
since the constraints in are a relaxation of the constraints of , the solution @xmath147 is also a feasible solution of .
based on proposition [ propositionlp1 ] we know that to any @xmath27 at least one of the following cases applies : 1 .
@xmath142 i.ethe constraint @xmath148 is tight .
notice that for @xmath167 satisfying the constraint @xmath168 we have , due to the set @xmath169 containing the set @xmath170 , @xmath171 .
according to proposition [ propositionlp1 ] we know that @xmath172 and @xmath145 . constraint ( 2 ) of _ fr - dist _ guarantees that @xmath173 . since we can apply case ( i ) for @xmath174 , we have @xmath175 , which yields @xmath176 .
given any instance of the facility location problem , we know that the optimal solution to the lp is bounded by @xmath41 and that , by proposition [ propositionlp2 ] , the solution to for the same instance is at most by a factor of @xmath177 larger . since the problem instance was chosen arbitrarily , the theorem s statement follows .
in this section we are only concerned with a single _ phase _ of our algorithm ( i.ethe loop in which algorithm 3 and 4 are invoked and which establishes fact @xmath21 ) .
we show that it terminates after @xmath114 rounds with high probability .
to simplify notation , @xmath6 and @xmath7 do not refer to all facilities respall clients ( as was the case before ) .
instead , let @xmath6 represent the set of all facilities that are in the _ currently - paid _ state and @xmath7 denote the set of all _ not connected _ clients contributing to a _ currently - paid _ facility . all the other clients and facilities effectively do not take part in the execution of algorithm 3 and 4 in the considered phase and thus can be ignored .
we consider the bipartite graph @xmath178 .
there is an edge in @xmath179 between @xmath8 and @xmath180 if and only if client @xmath13 contributes to @xmath14 ( i.e. @xmath181 ) .
furthermore , we assume that there are no isolated nodes ( i.e@xmath182 for all @xmath183 ) .
let @xmath184 .
by @xmath185 we denote the @xmath186 , where there is an edge @xmath187 between two facilities @xmath188 if and only if they share a common client in @xmath11 .
we use @xmath189 and @xmath190 to denote the degree and neighborhood of a node in @xmath11 respectively .
if the neighborhood includes the node itself , we write @xmath191 .
similarly , we use @xmath192 , @xmath193 , and @xmath194 to denote the corresponding properties in @xmath195 . in each execution of algorithm 3 and 4 the set @xmath6 of _ currently - paid _
facilities shrinks ( they change their status to _ closed _ or _ open _ ) .
also , with every facility changing its state from _ currently - paid _ to _ open _ clients that are connected to those newly opened facilities are removed from @xmath7 .
removing these clients form @xmath7 can potentially cause _ currently - paid _ facilities to change their state to _ closed _ ( they lose contributing clients and their opening costs are not paid for anymore ) . since it can not be guaranteed that a _ currently - paid _ facility which loses a contributing client is not longer fully paid for , applying a distributed mis ( maximal independent set ) algorithm ( e.gluby s mis algorithm @xcite ) on @xmath195 does not help to solve our problem : a facility that loses a contributing client ( i.ea facility that is not part of the mis ) might still be fully paid for . if such a facility is not opened in the current phase ( i.e , ( iii ) of fact 1 does not hold )
, clients contributing to it will raise their @xmath27 values in the next phase , which in turn could violate the constraints of the lp . since shrinking @xmath6 by removing _
currently - paid _ facilities that are not paid for only decreases the runtime of algorithm 3 and 4 , we will assume ( in order to consider the worst case ) that facilities never change their status from _ currently - paid _ to _ closed _ as the result of clients changing their state to _ connected _
( i.e , losing contributing clients ) . [ [ a - simplified - problem - description . ] ] a simplified problem description .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + in order to alleviate describing the shrinking process of the set of _ currently - paid _ facilities , we simplify the analysis of algorithm 3 and 4 by omitting the information about the values of @xmath24 and @xmath30 and simulating the distributed execution of the algorithms by the algorithm facilityselect .
its input is the bipartite graph @xmath11 and it possesses global knowledge .
@xmath196 ( _ subsequently , determine @xmath197 such that @xmath198 for all @xmath199 ) _ ) + generate a uniformly distributed random value @xmath200 $ ] + add @xmath14 to @xmath201 + eliminate @xmath202 from @xmath11 + as the execution of facilityselectprogresses , @xmath6 , @xmath7 , @xmath179 , and @xmath4 change .
we denote these sets and @xmath4 in round @xmath203 as @xmath204 , @xmath205 , @xmath206 and @xmath207 .
however , most of the time we consider only the effect of a single iteration of the while loop on the graph . in these cases ,
we omit the time parameter and simply use @xmath6 , @xmath7 , @xmath179 , and @xmath4 to refer to the corresponding values in the considered iteration . [
[ time - required - to - shrink - the - bipartite - graph . ] ] time required to shrink the bipartite graph .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + this paragraph analyzes the time required to remove all nodes from @xmath11 .
it is easy to see that a fixed facility @xmath208 is selected with probability @xmath209 by this strategy .
first , we show that , in expectation , facilityselectremoves at least @xmath210 clients and at least @xmath211 edges in a single iteration of the while loop .
[ prop : lower_bound_on_expected_eliminated_clients ] the expected number of clients removed by facilityselectin a single iteration of the while loop is at least @xmath210 . for a facility @xmath208 ,
let @xmath212 denote a random variable counting the number of clients @xmath213 removed because of @xmath14 being added to @xmath201 or not .
that is , @xmath214 since @xmath201 is an independent set in @xmath195 , the total number of clients removed in one iteration can be described by the random variable @xmath215 . using the inequality @xmath216 ,
we compute : @xmath217 [ prop : lower_bound_on_expected_eliminated_edges ] the expected number of edges removed by facilityselectin a single iteration of the while loop is at least @xmath211 .
first note that , since by lemma [ prop : lower_bound_on_expected_eliminated_clients ] the expected number of clients eliminated is at least @xmath210 and there are no isolated clients , the expected number of edges removed is at least @xmath210 .
it remains to show that the expected number of edges removed is at least @xmath218 . for a facility @xmath208 ,
let us define random variables @xmath219 counting the number of edges removed because of @xmath14 being added to @xmath201 or not .
that is , @xmath220 since @xmath201 is an independent set in @xmath195 , the total number of edges removed by strategy facilityselectcan be described by the random variable @xmath221 defined as @xmath222 . using the inequality @xmath223 ,
we compute : @xmath224 as an corollary , we get that , in expectation , the number of edges is reduced by at least its square root : [ cor : lower_bound_on_expected_eliminated_edges ] the expected number of edges removed by facilityselectin a single iteration of the while loop is at least @xmath225
. lemma [ prop : lower_bound_on_expected_eliminated_edges ] and the equality @xmath226 imply @xmath227 .
if the number of edges is greater or equal @xmath228 , the expected number of _ clients _ removed by facilityselectin one iteration of the while loop is at least @xmath229 . if it is smaller or equal @xmath228 , the expected number of removed _ edges _ is at least @xmath230 .
for the first statement , we can bound the expected number of removed clients with the help of lemma [ prop : lower_bound_on_expected_eliminated_clients ] as follows : @xmath231 the second statement follows immediately from corollary [ cor : lower_bound_on_expected_eliminated_edges ] .
we call a round @xmath203 _ edge - heavy _ , if in this round the number of edges is at least @xmath232 .
otherwise , we refer to it as _ edge - light_. [ client_constant_fraction ] for an arbitrary round @xmath203 , consider the effect of several iterations of facilityselect s while loop starting at round @xmath203 .
a. after @xmath233 ( not necessarily consecutive ) _ edge - heavy _ rounds ,
at least @xmath234 clients are removed in expectation .
b. after @xmath235 ( not necessarily consecutive ) _ edge - light _
rounds , at least @xmath236 edges are removed in expectation .
we prove only the first statement , since the second can be proven analogously .
let @xmath237 denote a random variable counting the number of clients removed during @xmath233 edge - heavy rounds .
if @xmath238 , the number of clients in each of the considered rounds was at least @xmath239 . thus , by lemma
[ prop : lower_bound_on_expected_eliminated_clients ] , in each of these @xmath233 rounds at least @xmath240 clients were removed in expectation . if @xmath241 , the number of removed clients is trivially lower bounded by @xmath239 .
we get : @xmath242&\geq\frac{{\left\lvertc_t\right\rvert}}{2}\cdot pr\left(y\geq\frac{{\left\lvertc_t\right\rvert}}{2}\right)+\frac{1}{2}\sqrt{n_t}\cdot\sqrt{\frac{{\left\lvertc_t\right\rvert}}{2}}\cdot pr\left(y<\frac{{\left\lvertc_t\right\rvert}}{2}\right)\\ & \geq\frac{{\left\lvertc_t\right\rvert}}{2\sqrt{2}}\cdot pr\left(y\geq\frac{{\left\lvertc_t\right\rvert}}{2}\right)+\frac{{\left\lvertc_t\right\rvert}}{2\sqrt{2}}\cdot pr\left(y<\frac{{\left\lvertc_t\right\rvert}}{2}\right ) = \frac{{\left\lvertc_t\right\rvert}}{2\sqrt{2 } } \enspace .
\end{split}\ ] ] [ client_edge_constant_probability ] for an arbitrary round @xmath203 , consider the effect of several iterations of facilityselect s while loop starting at round @xmath203 .
a. [ client_constant_probability ] the probability of removing at least @xmath243 clients after @xmath233 _ edge - heavy _
rounds is at least a constant @xmath244 .
b. [ edge_constant_probability ] the probability of removing at least @xmath245 edges after @xmath235 _ edge - light _
rounds is at least a constant @xmath244 .
we prove only the first statement .
the second can be proven analogously . by lemma [ client_constant_fraction ]
@xmath234 clients are removed after @xmath233 _ edge - heavy _
rounds in expectation .
let @xmath246 define a random variable counting the number of clients removed after @xmath233 _ edge - heavy _ rounds .
now define @xmath247 to be the probability that more than @xmath243 clients are removed .
we can bound @xmath248 $ ] in the following way : @xmath249 \leq \mu n + ( 1-\mu ) \frac{1}{4\sqrt{2}}{\left\lvertc_t\right\rvert}\ ] ] solving this inequality for @xmath247 yields @xmath250 , where @xmath54 is a positive constant .
we are now ready to prove our main theorem .
algorithm facilityselectterminates with high probability after @xmath251 iterations of the while loop . by lemma [ client_edge_constant_probability ] , starting at iteration @xmath203 we lose at least a constant fraction of either the clients or the edges with a non - zero , constant probability after at most @xmath252 rounds .
partition the number of iterations done by facilityselectinto phases of @xmath253 rounds .
let @xmath212 denote a binary random variable indicating whether the @xmath14-th phase was _ good _ ( we lost at least the above mentioned constant fraction of either clients or edges ) .
we need about @xmath254 good rounds to ensure that the problem size has become @xmath52 .
since we have @xmath255 for some positive constant @xmath54 , we get for @xmath256 : @xmath257\geq2\log(n ) \enspace.\ ] ] that is , in expectation it takes about @xmath258 phases until facilityselectterminates . applying a standard chernoff bound yields : @xmath259\right\rvert}>\frac{e[x]}{2}\right ) < 2^{-4e[x]/3 } \enspace.\ ] ] in other words ,
facilityselectterminates with high probability after at most @xmath260 steps .
we presented a parallel execution of a greedy sequential algorithm for the facility location problem and showed that we can preserve its approximation factor .
however , there are other sequential facility location algorithms based on the greedy approach introduced by jain et althat yield ( better ) approximation factors of 1.61 @xcite and 1.52 @xcite
. it might be possible to execute them in the @xmath0 model with a sublinear running time using the technique we introduced in this paper .
when our algorithm chooses which facilities should be opened , it bases its decision solely on the information provided by the bipartite graph @xmath11 used in section 3 and all other information ( client contribution , distance , facility opening costs ect . )
is abstracted away ( section 3 provides reasons for this abstraction ) . at this level of abstraction
the lower bound of solving the problem is @xmath261 : consider a bipartite graph with @xmath261 facilities with degree of @xmath262 , and @xmath263 clients with degree @xmath23 .
edges are chosen in such a way that any two facilities are in hop distance of 2 of each other .
it would be interesting to know in what way the entire available information can be used to improve the runtime and whether a poly - logarithmic runtime is possible with the same approximation factor .
if @xmath264 the lemma obviously holds . thus , let us assume the opposite .
let @xmath203 be the phase when our algorithm connects client @xmath125 to some facility @xmath93 and @xmath265 the phase @xmath93 was opened .
facility @xmath93 was either opened before @xmath125 was connected to it ( @xmath266 ) or opened exactly in the same phase client @xmath125 was connected to it ( @xmath267 ) .
since @xmath268 , we know that client @xmath13 was connected in phase @xmath269 . at time
@xmath270 the contribution of client @xmath13 was not enough to cover the distance between @xmath13 and @xmath93 , since otherwise @xmath13 would be connected in phase @xmath265 with @xmath93 .
this means that once the contribution of @xmath13 reaches @xmath271 , client @xmath13 will be connected to @xmath93 .
since the contribution is increased by multiplying the current @xmath147 value with @xmath49 , it can not become greater than @xmath272 , hence @xmath273 .
thanks to the triangle inequality we have @xmath274 . to prove by contradiction , we assume that the inequality does not hold . thus , we have @xmath275 . since we ordered the clients according to their @xmath147 value , we know that for @xmath276 we have @xmath277
let @xmath203 be the phase client @xmath13 was connected to a facility . by the assumption facility @xmath14
is fully paid for ( i.e @xmath14 s current payment is @xmath278 ) in phase @xmath266 .
there must be at least one client @xmath174 , with @xmath279 and @xmath280 which was connected to @xmath14 in a phase @xmath281 . for this client @xmath174
, we have @xmath282 , since the algorithm will stop increasing @xmath283 when @xmath174 is connected to a fully paid facility ( see fact 1 ) . | we present a randomized distributed approximation algorithm for the metric uncapacitated facility location problem .
the algorithm is executed on a bipartite graph in the @xmath0 model yielding a @xmath1 approximation factor , where @xmath2 is an arbitrary small positive constant .
it needs @xmath3 communication rounds with high probability ( @xmath4 denoting the number of facilities and clients ) . to the best of our knowledge
, our algorithm currently has the best approximation factor for the facility location problem in a distributed setting .
it is based on a greedy sequential approximation algorithm by jain et al(j .
acm 50(6 ) , pages : 795 - 824 , 2003 ) .
the main difficulty in executing this sequential algorithm lies in dealing with situations , where multiple facilities are eligible for opening , but ( in order to preserve the approximation factor of the sequential algorithm ) only a subset of them can actually be opened .
note that while the presented runtime bound of our algorithm is `` with high probability '' , the approximation factor is not `` in expectation '' but always guaranteed to be @xmath1 .
thus , our main contribution is a sublinear time selection mechanism that , while increasing the approximation factor by an arbitrary small additive term , allows us to decide which of the eligible facilities to open . |
manipulation of atoms by light becomes possible due to the dipole forces which are well described by the semiclassical model with quantum description of internal atomic transitions induced by a near resonant laser field and classical description of their center - of - mass motion @xcite .
for the first time , the ideas to trap and channel cold atoms with the help of standing laser waves ( slw ) have been proposed by v. letokhov and his co - workers @xcite .
the ability of a slw to deflect , channel and split atomic beams @xcite has been used for a variety of applications including atom microscopy , interferometry , isotope separation and optical lithography @xcite .
lasers can be used to manipulate atomic trajectories to create atomic analogues of such familiar optical phenomena as focusing of light , beam splitting and light scattering .
it is remarkable that now we are able to reverse the roles of light and matter from their familiar roles .
the semiclassical description , used in this paper , is similar to the geometric optics limit in conventional optics .
atomic trajectories play the role of light rays with the slw being a light mask . in the present paper
we intend to demonstrate theoretically and numerically that adjusting in an experiment only one parameter , the detuning between the frequencies of a working atomic transition and the slw , one can explore a variety of the regimes of the atom - laser interaction to focus , split and scatter atomic beams . near the atom - field resonance , where the interaction between the internal and external atomic degrees of freedom is intense , there is a possibility to create conditions for chaotic scattering of atoms @xcite without any additional efforts like a slw modulation .
it becomes possible due to the peculiarities of the dipole force in a near - resonant optical lattice @xcite .
a beam of two - level atoms in the @xmath1 direction crosses a slw laser field with optical axis in the @xmath2 direction .
the laser beam has the gaussian profile @xmath3 $ ] with @xmath4 being the @xmath5 radius at the laser beam waist .
the characteristic length of the atom - field interaction is supposed to be @xmath6 because the light intensity at @xmath7 is two orders of magnitude smaller than the peak value .
the longitudinal velocity of atoms , @xmath8 , is much larger than their transversal velocity @xmath9 and is supposed to be constant .
thus , the spatial laser profile may be replaced by the temporal one .
the hamiltonian of a two - level atom in the one - dimensional slw can be written in the frame rotating with the angular laser frequency @xmath10 as follows : @xmath11 \left(\hat\sigma_-+\hat\sigma_+\right)\cos{k_f x_a } -\frac{i\hbar\gamma}{2}\hat\sigma_+\hat\sigma_- , \label{ham}\ ] ] where @xmath12 are the pauli operators for the internal atomic degrees of freedom , @xmath13 and @xmath14 are the classical atomic position and momentum , @xmath15 , @xmath16 , and @xmath17 are the decay rate , the atomic transition and maximal rabi frequencies , respectively .
the simple wave function for the electronic degree of freedom is @xmath18 , where @xmath19 and @xmath20 are the complex - valued probability amplitudes to find the atom in the excited , @xmath21 , and ground , @xmath22 , states , respectively . in the semiclassical approximation ,
atom with quantized internal dynamics is treated as a point - like particle to be described by the hamilton
schrdinger equations of motion written for the real and imaginary parts of the probability amplitudes @xmath23(ab+\alpha \beta)\sin x , \\ \dot a = \frac12 ( \omega_rp^2 - \delta)\alpha - \frac12 \gamma a - \exp[-(\tau-\frac{3}{2 } \sigma_{\tau})^2/\sigma^2_{\tau}]\beta\cos x , \\
\dot \alpha = -\frac12 ( \omega_rp^2 - \delta)a - \frac12 \gamma \alpha + \exp[-(\tau-\frac{3}{2 } \sigma_{\tau})^2/\sigma^2_{\tau}]b\cos x , \\ \dot b = \frac12 ( \omega_rp^2 + \delta)\beta - \exp[-(\tau-\frac{3}{2 } \sigma_{\tau})^2/\sigma^2_{\tau}]\alpha \cos x , \\ \dot \beta = -\frac12 ( \omega_rp^2 + \delta)b+ \exp[-(\tau-\frac{3}{2 } \sigma_{\tau})^2/\sigma^2_{\tau}]a\cos x , \end{gathered } \label{mainsys}\ ] ] where @xmath24 and @xmath25 are scaled atomic center - of - mass position and transversal momentum , respectively and dot denotes differentiation with respect to the dimensionless time @xmath26 .
the recoil frequency , @xmath27 , the atom - laser detuning , @xmath28 , the decay rate @xmath29 , and the characteristic interaction time , @xmath30 , are the control parameters .
let us introduce instead of the complex - valued probability amplitudes @xmath31 and @xmath32 the following real - valued variables : @xmath33 where @xmath34 and @xmath35 are synchronized and quadrature components of the atomic electric dipole moment , respectively , and @xmath1 is the atomic population inversion . in the absence of any losses ( @xmath36 ) , eqs .
( [ mainsys ] ) can be cast in the form @xmath37\sin x , \quad \dot u=\delta v,\\ \dot v=-\delta u+2\exp[-(\tau-\frac{3}{2 } \sigma_{\tau})^2/\sigma^2_{\tau } ] z\cos x , \quad \dot z=-2\exp[-(\tau-\frac{3}{2 } \sigma_{\tau})^2/\sigma^2_{\tau } ] v\cos x. \end{gathered } \label{hamsys}\ ] ] the system ( [ hamsys ] ) has two integrals of motion , namely , the total energy @xmath38 and the length of the bloch vector , @xmath39 , whose conservation follows immediately from eqs .
( [ uvz_def ] ) .
equations ( [ hamsys ] ) constitute a nonlinear hamiltonian autonomous system with two and half degrees of freedom which , owing to the two integrals of motion , move on a three - dimensional hypersurface with a given energy value @xmath40 . in general , motion in a three - dimensional phase space in characterized by a positive lyapunov exponent @xmath41 , a negative exponent equal in magnitude to the positive one , and zero exponent @xcite .
the maximal lyapunov exponent characterizes the mean rate of the exponential divergence of initially close trajectories and serves as a quantitative measure of dynamical chaos in the system .
the values of the maximal lyapunov exponent in dependence on the detuning , the recoil frequency and the initial atomic momentum have been computed in refs .
@xcite .
there are different regimes of the center - of - mass motion along the slw optical axis @xcite . in dependence on the initial conditions and the values of the control parameters
, atoms may oscillate in a regular or a chaotic way in wells of the optical potential or move ballistically over its hills with regular or chaotic variations of their velocity .
chaotic motion with a positive value of the maximal lyapunov exponent becomes possible in a narrow range of the detuning values , @xmath42 @xcite . at @xmath43 ,
the synchronized electric - dipole component , @xmath34 , becomes a constant .
that implies the additional integral of motion in the hamiltonian version ( [ hamsys ] ) of eqs .
( [ mainsys ] ) and the regular motion with zero maximal lyapunov exponent . far off the resonance , at @xmath44 ,
the motion is regular both in the trapping and ballistic modes .
it is remarkable that there is a specific type of motion , chaotic walking in a deterministic optical potential , when atoms can change the direction of motion alternating between flying through the slw and being trapped in its potential wells .
we would like to stress that the local instability produces chaotic center - of - mass motion in a rigid slw without any modulation of its parameters .
chaotic walking occurs due to the specific behavior of the bloch - vector component of a moving atom @xmath34 whose shallow oscillations between the slw nodes are interrupted by sudden jumps with different amplitudes while atom crosses each node of the slw @xcite .
it looks like a random like shots happened in a fully deterministic environment .
it follows from the second equation in the set ( [ hamsys ] ) that those jumps result in jumps of the atomic momentum while crossing a node of the slw .
if the value of the atomic energy is close to the separatrix one , the atom after the corresponding jump - like change in @xmath45 can either overcome the potential barrier and leave a potential well or it will be trapped by the well , or it will move as before . the jump - like behavior of @xmath34 is the ultimate reason of chaotic atomic walking along a rigid sw .
the total atomic energy ( [ h ] ) consists of the kinetic one , @xmath46 , and the potential one , @xmath47 .
the optical potential changes its depth in course of time . averaging over fast oscillations of the internal atomic variables
, we get the averaged potential @xmath48 that can be used to explain why atoms move in such or another way . at small detunings @xmath49 ,
the potential is approximately @xmath50 .
if @xmath51 , then the atom will move ballistically .
this occurs if the initial atomic momentum , @xmath52 , satisfies to the condition @xmath53 .
if the initial conditions are chosen to give @xmath54 , the corresponding atoms with @xmath55 are expected to move chaotically at the appropriate values of @xmath56 . in this section
we demonstrate how to focus and split atomic beams crossing a gaussian laser beam by varying only one of the control parameters , the atom - field detuning @xmath56 .
firstly , we perform simulation with a negligible probability of spontaneous emission and solve the hamiltonian equations of motion ( [ hamsys ] ) at comparatively large value of the detuning , @xmath57 . to be concrete we take as an example calcium atoms with the working intercombination transition @xmath58 at @xmath59 nm , the recoil frequency @xmath60 khz , and the lifetime of the excited state @xmath61 ms .
taking the maximal rabi frequency to be @xmath62 hz , the radius of the laser beam @xmath63 cm , and the mean longitudinal velocity @xmath64 m / s , the interaction time is estimated to be @xmath65 ms , longer than the atomic lifetime .
the normalized recoil frequency is @xmath66 and the normalized characteristic time is @xmath67 .
trajectories for 50 calcium atoms to be prepared in the ground states ( @xmath68 , @xmath69 ) with the same initial momentum , @xmath70 , and initial positions in the range @xmath71 are shown in fig .
[ fig1 ] . in units of the optical wavelength , @xmath72 ,
this range is @xmath73 .
the focusing occurs at those moments of time when the average transverse momentum in the atomic beam is approximately equal to zero .
if one turns off the laser at one of these moments , it becomes possible to reduce the beam width practically in ten times .
the reason of focusing is simple .
it is well known @xcite that at positive blue detunings atoms are attracted to the nodes of the slw where the minima of the optical potential are situated at @xmath74 .
the first node the atoms reach at @xmath75 is situated at @xmath76 .
the initial kinetic atomic energy , @xmath77 , is not enough to overcome the potential barrier whose depth can be estimated to be @xmath78 because the simulation gives @xmath79 and @xmath80 .
so , all the atoms in the beam oscillate in the first potential well in the @xmath2-direction around the first node . the initial width of the beam , @xmath81 , is gradually reduced because in course of time the atoms with initial negative positions catch up with the ones with positive @xmath82 near the first turning point where the average beam momentum is close to zero .
the time interval of the atomic interaction with the slw field is estimated to be @xmath83 .
so , the atoms leave the potential well after that time and move freely ( see fig . [ fig1 ] ) . in order to take into account spontaneous emission
we use the standard stochastic wave - function technique @xcite for solving eqs .
( [ mainsys ] ) . the integration time is divided into a large number of small time intervals @xmath84 . at the end of the first interval , @xmath85 ,
the probability of spontaneous emission , @xmath86 , is computed and compared with a random number , @xmath87 , from the interval @xmath88 $ ] . if @xmath89 , then one prolongs the integration but renormalizes the state vector in the end of the first interval at @xmath90 : @xmath91 and @xmath92 with @xmath93 , @xmath94 .
its momentum in the @xmath2 direction changes for a random number from the interval @xmath88 $ ] due to the photon recoil effect , and the next time step commences .
we simulate lithium atoms with the relevant transition @xmath95 , the corresponding wavelength @xmath96 nm , recoil frequency @xmath97 khz , and the decay time @xmath98 s. with the maximal rabi frequency @xmath99 mhz and the radius of the laser beam @xmath100 cm one gets @xmath101 , @xmath67 , and @xmath102 . trajectories for 50 spontaneously emitting atoms under the same conditions as in fig .
[ fig1]a are shown in fig .
as expected , spontaneous emission destroys in part the effect of focusing .
however , the atoms move more or less coherently because spontaneous emission events are comparatively rare at @xmath57 . in both the cases .
the atomic position @xmath103 is in units of the optical wavelength.,title="fig:",scaledwidth=40.0% ] in both the cases .
the atomic position @xmath103 is in units of the optical wavelength.,title="fig:",scaledwidth=40.0% ] .
, title="fig:",scaledwidth=40.0% ] . , title="fig:",scaledwidth=40.0% ] the other effect , we would like to demonstrate with atomic beams crossing the slw , is a splitting of the beam . to do
this one needs to choose such the value of the detuning in order that some atoms in the beam would be trapped in the first well of the optical potential but another ones could overcome the barrier and leave that well .
it is possible to split atomic beams as at positive and negative values of the detuning . as an example
, we demonstrate in fig .
[ fig2 ] the effect of splitting at @xmath104 for atoms without and with spontaneous emission .
it is seen that spontaneous emission changes slightly the effect because a few atoms may leave the potential wells due to random recoils .
the ability of blue and red detuned lasers to attract atoms to the nodes and antinodes of the slw , respectively , can be used to create periodic structures composed of atoms deposited on substrates in the process of optical nanolithography @xcite . to simulate a real experiment we consider a beam with @xmath105 calcium atoms with the initial gaussian distribution ( with the rms @xmath106 and the average values @xmath107 and @xmath70 ) and
compute their distribution against the slw at a fixed moment of time @xmath108 . the bunching of atoms at the slw nodes at @xmath109 ( blue detuning ) is shown in fig .
[ fig3]a where the atomic density , @xmath110 , is plotted along the optical axis @xmath103 at @xmath108 . the same effect , but with the atoms bunching around the slw antinodes ( red detuning , @xmath111 ) , is shown in fig .
[ fig3]b . in both the cases
we get a periodic atomic relief with the period @xmath112 the width of which is restricted by the time the atoms interact with the gaussian laser beam . calcium atoms around ( a ) the slw nodes ( blue detuning , @xmath109 ) and ( b ) the antinodes ( red detuning , @xmath111 ) . the plot of atomic density @xmath110 at the fixed moment of time @xmath108.,title="fig:",scaledwidth=40.0% ] calcium atoms around ( a ) the slw nodes ( blue detuning , @xmath109 ) and ( b ) the antinodes ( red detuning , @xmath111 ) . the plot of atomic density @xmath110 at the fixed moment of time @xmath108.,title="fig:",scaledwidth=40.0% ] the problem we consider resembles the scattering process with particles entering an interaction region along completely regular trajectories and leaving it along asymptotically regular trajectories .
it is known from many studies in celestial mechanics , fluid dynamics and other disciplines that under certain conditions the motion inside the interaction region may have features that are typical for dynamical chaos , ( homoclinic and heteroclinic tangles , fractals , strange invariant sets , positive finite - time lyapunov exponents , etc . )
although the particle s trajectories are not chaotic in a rigorous sense because chaos is strictly defined as an irregular motion over infinite time .
it has been found @xcite that transient hamiltonian chaos in the interaction region occurs due to existence of , at least , one nonattractive chaotic invariant set consisting of an infinite number of localized unstable periodic orbits and aperiodic orbits .
this set possesses stable and unstable manifolds extending in the phase space into the regions of regular motion .
the particles with the initial positions close to the stable manifold follow the chaotic - set trajectories for a comparatively long time , then deviate from them , and leave the interaction region along the unstable manifold .
it is the common mechanism of chaotic scattering that in our problem causes the chaotic walking of atoms along the slw .
calcium atoms at @xmath108 in ( a ) the real and ( b ) momentum space under the conditions of chaotic scattering at @xmath113.,title="fig:",scaledwidth=40.0% ] calcium atoms at @xmath108 in ( a ) the real and ( b ) momentum space under the conditions of chaotic scattering at @xmath113.,title="fig:",scaledwidth=40.0% ] calcium atoms at @xmath108 over the phase plane in the regimes of ( a ) chaotic ( @xmath113 ) and ( b ) regular scattering ( @xmath114).,title="fig:",scaledwidth=40.0% ] calcium atoms at @xmath108 over the phase plane in the regimes of ( a ) chaotic ( @xmath113 ) and ( b ) regular scattering ( @xmath114).,title="fig:",scaledwidth=40.0% ] in fig .
[ fig4]a we show the atomic position distribution at @xmath108 in the regime of the chaotic scattering at @xmath115 with @xmath116 calcium atoms .
this plot should be compared with fig .
[ fig3]a where the atomic position distribution is shown for regularly scattered atoms at @xmath109 .
first of all , the distribution of chaotically scattered atoms has a prominent pedestal and is much broader .
moreover , it has no such a periodic structure as shown in fig .
only the peaks around the first two slw nodes are prominent in fig .
[ fig4]a . the atomic position distribution in the momentum space in fig .
[ fig4]b is much broader than the one for regularly scattered atoms at @xmath109 ( not shown ) .
thus , we predict that under the conditions of chaotic scattering there should appear less contrast and more broadened atomic reliefs as compared to the case of regular scattering because a large number of atoms are expected to be deposited between the nodes as a result of chaotic walking along the slw axis .
the effect is expected to be more prominent under the coherent evolution but it seems to be observable with spontaneously emitting atoms as well .
the difference between chaotic and regular scattering of atoms at a rigid slw is especially prominent on the corresponding phase space portraits shown in fig .
[ fig5 ] where positions and momenta of @xmath116 calcium atoms are plotted at the fixed time moment .
we have simulated some geometric optics effects with atomic beams crossing a slw in the limit of long relaxation time and with spontaneous emission taken into account .
trajectories of spontaneously emitting atoms have been simulated with the help of the standard stochastic wave - function technique @xcite . it has been shown that by adjusting the detuning it is possible to focus , split and scatter atoms .
the effects have been explained by a coupling between external and internal atomic degrees of freedom .
the depth of the optical potential depends on the sign and value of the detuning . varying @xmath56
, one can create conditions for focusing , splitting and bunching the atoms .
it is remarkable that near the atom - field resonance we have found the new type of atomic diffraction at a slw without any modulation of its parameters that can be observed in real experiments
. that would be the prove of existence of the novel type of atomic motion , chaotic walking in a deterministic environment .
the effects found could be used in optical nanolithography to fabricate complex atomic structures on substrates .
we predict that experiments on the scattering of atomic beams at a slw can directly image chaotic walking of atoms along the slw . in a real experiment
the final spatial distribution can be recorded via fluorescence or absorption imaging on a ccd , commonly used methods in atom optics experiments yielding information on the number of atoms and the cloud s spatial size .
the other possibility is a nanofabrication where the atoms after the interaction with the sw are deposited on a silicon substrate in a high vacuum chamber . in this case
the spatial distribution can be analyzed with an atomic force microscope .
as to the momentum distribution , it can be measured , for example , by a time - of - flight technique @xcite .
the modern tools of atom optics enable to create narrow initial atomic distributions in position and momentum , reduce coupling to the environment and technical noise , create one - dimensional optical potentials , and to measure spatial and momentum distributions with high sensitivity and accuracy .
this work was supported by the russian foundation for basic research ( projects nos .
09 - 02 - 00358 and 09 - 02 - 01258 ) , by the integration grant from the far - eastern and siberian branches of the russian academy of sciences , and by the program `` fundamental problems of nonlinear dynamics '' . 99 a.p .
kazantsev , g.a .
ryabenko , g.i .
surdutovich , v.p .
yakovlev , _ phys .
_ , * 129 * , 75 ( 1985 ) .
letokhov , _ jetp lett _ , * 7 * , 272 ( 1968 ) [ _ pisma zhetf _ , * 7 * , 348 ( 1968 ) ] . v.i .
balykin , v.s .
letokhov , _ opt .
_ , * 64 * , 151 ( 1987 ) .
v. letokhov , _ laser control of atoms and molecules _ ( oxford university press , new york , 2007 ) .
e. arimondo , a. bambini , s. stenholm , _ phys .
_ , * 24 * , 898 ( 1981 )
adams , m. sigel , j. mlynek , _ phys .
_ , * 240 * , 143 ( 1994 ) .
t. sleator , t. pfau , v. balykin , o. carnal et al , _ phys .
_ , * 68 * , 1996 ( 1992 ) .
g. timp , r.e .
behringer , d.m .
tennant , j.e .
cunningham , _ phys .
_ , * 69 * , 1636 ( 1992 ) .
mcclelland , r.e .
scholten , e.c .
palm , r.j .
celotta , _ science _ , * 262 * , 877 ( 1993 ) .
s. v. prants , l.e .
konkov , _ jetp letters _ ,
* 73 * , 1801 ( 2001 ) [ _ pisma zhetf _ , * 73 * , 200 ( 2001 ) ] . s. v. prants , m. edelman , g. m. zaslavsky , _ phys .
e _ , * 66 * , art .
046222 ( 2002 ) .
prants , v.yu .
sirotkin , _ phys .
a _ , * 64 * , 033412 ( 2001 ) . v. yu .
argonov , s. v. prants , _ jetp _ , * 96 * , 832 ( 2003 ) [ _ zhetf _ , * 123 * , 946 ( 2003 ) ] .
argonov , s. v. prants , _
laser res .
_ , * 27 * , 360 ( 2006 ) v. yu .
argonov , s. v. prants , _ phys .
a _ , * 75 * , art .
063428 ( 2007 ) .
argonov , s. v. prants , _ phys .
a _ , * 78 * , art . 043413 ( 2008 ) .
konkov , s. v. prants , _ jetp letters _ , * 65 * , 833 ( 1997 ) [ _ pisma zhetf _ , * 65 * , 801 ( 1997 ) ] . h. j. carmichael , _ an open systems approach to quantum optics _
( berlin , springer , 1993 ) . j. dalibard , y. castin , k. mlmer , _ phys . rev .
_ , * 68 * , 580 ( 1992 ) .
r. dum , p. zoller , h. ritsch , _ phys .
a _ , * 45 * , 4879 ( 1992 ) .
p. gaspard , _ chaos , scattering and statistical mechanics _ , cambridge university press , cambridge ( 1998 ) .
mitchell , j.p .
handley , b. tighe , j.b .
delos , s.k .
knudson , _ chaos _ , * 13 * , 880 ( 2003 ) .
m. budyansky , m. uleysky , s. prants , _ physica d _ , * 195 * , 369 ( 2004 ) .
budyansky , m.yu .
uleysky , s.v .
prants , _ jetp _ , * 99 * , 1018 ( 2004 ) [ _ zhetf _ , * 126 * , 1167 ( 2004 ) ] . m. g. raizen , _ adv . at .
_ , * 41 * , 43 ( 1999 ) . | we report on theoretical and numerical study of propagation of atomic beams crossing a detuned standing - wave laser beam in the geometric optics limit .
the interplay between external and internal atomic degrees of freedom is used to manipulate the atomic motion along the optical axis by light . by adjusting the atom - laser detuning
, we demonstrate how to focus , split and scatter atomic beams in a real experiment . the novel effect of chaotic scattering of atoms at a regular near - resonant standing wave is found numerically and explained qualitatively .
some applications of the effects found are discussed .
[ sh ] * * [ cols="^ " , ] * s.v .
prants , v.o .
vitkovsky , l.e .
konkov * _ laboratory of nonlinear dynamical systems , + pacific oceanological institute of the russian academy of sciences , + 690041 vladivostok , russia , url : dynalab.poi.dvo.ru_ @xmath0corresponding author e - mail : [email protected] + * keywords : * atomic scattering , standing wave , optical nanolithography |
after the so - called second string revolution , string phenomenology has widen its subject beyond heterotic models .
one very exciting possibility came across a few years ago through the realization that d - branes intersecting at non - trivial angles @xcite allow for a neat way of breaking supersymmetry @xcite , generating four - dimensional chirality , and constructing in a bottom - up approach string models with a fully realistic matter and symmetry content @xcite .
a lot of effort has been put since in the study of the model building and phenomenology of such theories .
( see @xcite for the initial works and @xcite for a more detailed introduction and references . )
this intense dedication has resulted almost futile regarding the search of supersymmetric models .
the restrictions from ramond - ramond tadpole cancellation are so strict that when added to the requirement of the preservation of @xmath0 supersymmetry the resulting models are scarce and usually not fully realistic @xcite , with extra exotic matter in the spectrum .
( see however important progress in this direction in ref .
@xcite . )
given the fact that most of the realistic models presented so far are non - supersymmetric , a sensible question to ask is whether low scale non - fine - tuned models are phenomenologically allowed
. a necessary previous step , to which we devote the first half for the present review , is the investigation of the structure of flavor in these models .
the trivial structure of tree level yukawa couplings present in realistic models @xcite might seem discouraging at first sight . however the presence of flavor violating processes @xcite , both at the string and field theory levels , propagates through quantum corrections to give a rich , realistic structure of fermion masses and mixing angles when one loop contributions are taken into account .
armed with this realistic flavor structure we can study , in a quantitative unambiguous way , the phenomenological implications of low scale intersecting brane models .
the same flavor violating four - point amplitudes that give raise to a realistic fermion spectrum induce important tree - level contributions to meson oscillations and rare processes .
these are extremely well constrained experimentally , implying very stringent bounds on the string scale .
( incidentally , we should mention here that while tree level contributions to flavor violating processes decouple as the string scale increases , their contribution to the one loop yukawa couplings does not . ) in an attempt to be as comprehensive as possible and in order to minimize the possibility of fine - tuned situations in which the flavor violating contributions are small , we also discuss a new host of bounds on these models arising from flavor conserving amplitudes .
although the resulting constraints are much milder , still a lower bound on the string scale in the tens of tev region is obtained .
models with intersecting branes have a number of very interesting features such as the presence of chiral fermions , family replication and the possibility of constructing models with precisely the symmetries and the matter content of the standard model .
once a model has successfully passed such a coarse graining sift one has to more finely test it by for instance checking whether a realistic pattern of fermion masses and mixing angles can arise and what phenomenological predictions and restrictions it has .
two main ingredients enter the flavor structure of these models , one is the form of the yukawa couplings , the other the presence of tree level flavor changing four point amplitudes .
these two features are intimately related for , as we shall see , flavor violating operators are needed to generate a fully realistic pattern of fermion masses and mixings but at the same time , bounds on the string scale from the contribution of these operators to rare processes can not be computed unless the flavor structure is known . in order to make our arguments more specific , we shall concentrate on a particular model represented in fig .
[ figuratoros ] @xcite .
it corresponds to an orientifold compactification of type ii a theory with four stacks of d6-branes wrapping factorizable 3-cycles on the compact dimensions .
the compactified space is a factorizable 6-torus @xmath1 , and the orientifold projection is given by @xmath2 where @xmath3 is the world - sheet parity and @xmath4 is a reflection about the horizontal axis of each of the three 2-tori , @xmath5 we have denoted the coordinates of the tori by complex coordinates @xmath6 , @xmath7 , so the three boxes in the figure represent each 2 torus , with opposite edges being identified .
the branes have four extended dimensions plus three compactified ones .
therefore , they appear as just lines in each @xmath8 .
the net effect of the orientifold projection is to introduce mirror images of the branes in each @xmath8 ( in the plane running horizontally ) .
the matter content of the low energy spectrum consists of massless chiral 4-dimensional fermions plus generally massive scalars , both living at branes intersections and transforming as bi - fundamentals of the corresponding gauge groups .
the masses of the latter depend on the particular brane configuration ( angle between branes ) and can be considered as the superpartners of the fermions ( for specific values of the angles in which they become massless , some @xmath0 supersymmetry will be preserved by the corresponding intersection ) .
gauge bosons live in the world - volume of the branes , corresponding to unitary ( orthogonal or simplectic groups are also possible for orientifold compactifications ) gauge groups .
our particular model contains at low energies just the particle content and symmetries of the mssm . in order to get that
, the model contains four stacks of d6-branes , called _ baryonic _ ( a ) , _ left _ ( b ) , _ right _ ( c ) , and _ leptonic _ ( d ) .
three of the dimensions of each d6-brane wrap a 1-cycle on each of the three 2-tori , with wrapping numbers denoted by @xmath9 , _ i.e. _ the stack @xmath10 wraps @xmath11 times the horizontal dimension of the @xmath12th torus and @xmath13 times the vertical direction .
we have to include for consistency their orientifold images with @xmath14 wrapping numbers .
the number of branes in each stack , their wrapping numbers and the gauge groups they give rise to are shown in table [ data : branes : model ] and a subset of them , together with some of the relevant moduli , are displayed in fig .
[ figuratoros ] . a crucial feature that governs the whole flavor structure of these models is the fact that different families as well as the higgs boson live at separate points in the compact dimensions .
yukawa interactions then correspond to instantonic contributions that are therefore expected to be proportional to the exponential of the relevant area connecting the three vertices .
a more detailed study of yukawa couplings , using calibrated geometry @xcite , and confirmed later by a proper string calculation using conformal field theory techniques @xcite , showed that when the compact space is a factorizable torus and the branes wrap factorizable cycles , the relevant area is the sum of the _ projected _ areas of the triangle over each sub - torus . the final result , including the quantum part reads @xmath15 where we have neglected the presence of non - zero @xmath16 field and wilson lines and @xmath17 is the string coupling , @xmath16 is the euler beta function , @xmath18 runs over the three tori , @xmath19 and @xmath20 are the angles at the fermionic intersections , @xmath21 runs over all possible triangles connecting the three vertices on each of the three tori ( there is an infinite number of them due to the toroidal periodicity ) and @xmath22 is the projected area of the @xmath23th triangle on the @xmath12th torus .
another important feature of the model is the fact that family _ separation _ occurs at a different torus for each chirality . as can be seen in fig .
[ figuratoros ] , left handed fermions are split apart in the second torus ( called hereafter _ left _
torus ) whereas they are localized at the same point in the third torus ( the _ right _ one ) .
the opposite happens for the right handed fields .
this fact , together with the previously mentioned that the couplings are proportional to the exponential of minus the _ projected _ areas , result in a trivial , factorizable form of the yukawa couplings at tree level @xmath24 the resulting spectrum is clearly unrealistic , with one massive and two massless generations . nonetheless we will shortly see that the same feature of fermion splitting in the compact dimensions originates flavor violating four - point amplitudes that will contribute at loop level to give a non - trivial , fully realistic structure to the yukawa couplings .
it has been known for long that models with split fermions suffer from flavor changing neutral current problems arising from the family non - universal couplings of the gauge boson kk modes @xcite .
a detailed study of the four - point amplitudes @xcite in these models uncovered new sources of flavor violation mediated by string instantons @xcite .
these can be chirality preserving , whose contribution is again proportional to the exponential of the area of the corresponding quadrangle ( with a non - zero area in only one torus for the leading contribution ) or chirality changing , with a much richer structure and relevance regarding the generation of non - trivial one loop yukawa couplings .
it is the latter that we turn our attention to now .
we are interested in the following chirality changing four - point amplitudes ( for concreteness we consider here quark fields as external particles ) @xmath25 at the field theory level , such amplitude can be originated by the exchange of the higgs boson .
this process is indeed reproduced in the string calculation , together with other processes , purely stringy in origin , that will be crucial to generate a non - trivial fermionic spectrum . for the sake of the discussion we concentrate for the moment on one particular sub - torus .
the relevant amplitude corresponds to a four - sided polygon with two left handed and two right handed quarks as vertices .
there are two possibilities for such a polygon , displayed in fig .
[ crossvsnoncross ] , one is obtained by adding together two triangular areas joined by the higgs vertex ( green area in the figure ) , the string calculation agrees with the intuition , giving a contribution proportional to the product of the two yukawa couplings ( one per triangle ) and with a higgs pole structure , @xmath26 where we have shown the @xmath27-channel exchange for the sake of the example .
the second possibility , represented in blue in the figure , lacks a field theory counterpart . in the case
that there is a non - zero area in just one torus , this contribution is again proportional to the exponential of the area swept out by the string connecting the four vertices , this time without running through a higgs vertex and therefore being suppressed by the string scale instead of having a higgs pole like the previous contribution .
two important effects are originated from this kind of diagrams , the first is already apparent now . even in the case of non - zero area in just one torus , although the amplitude is factorizable , it is not proportional to the yukawa couplings , therefore introducing new , yet too trivial to generate a fully realistic spectrum , flavor structure in the game . the second feature , this time enough to realize a realistic fermion spectrum happens when there are non - zero areas in more than one torus . in that case , as we shall see in a moment , factorization of the amplitude is lost and fully non - trivial new flavor structure is generated .
the technical reason for this non - factorization is the following . in order to perform the string calculation one maps the world - sheet into the upper complex plane , with the vertices in the real axis .
three of the vertices are conventionally fixed at positions 0 , 1 and @xmath28 using @xmath29 invariance whereas one has to integrate over the position of the last one , that is situated at @xmath30 .
the classical contribution is then obtained by minimizing the action with respect to this parameter @xmath31 and performing a saddle point approximation .
the crucial point is that there is only _
parameter @xmath31 for _ all _ the three tori and although the minimization of the action for each torus gives the projected area in that torus , the value of @xmath32 is usually not the same on the three of them and therefore the minimum of the total action does not correspond to the separate minimum of each sub - factor . apart from the trivial cases in which the areas are equal up to rescaling in all tori or
there is a non - zero area in just one torus , there are two ( related ) cases in which there is still factorization .
one corresponds to three point amplitudes that give rise to yukawa couplings , they are proportional to the projected areas and therefore factorize because we can fix the three vertices and there is not any extra free parameter that could make a difference between the different tori .
the second case is chirality flipping amplitudes mediated by the higgs field . in that case , the minimum of the action corresponds to @xmath33 , which is of course common to all tori .
we have seen that the tree level yukawa couplings are factorizable and therefore too trivial for a realistic spectrum .
chirality and flavor changing amplitudes have been found of three different classes flavor wise , one is mediated by the higgs boson , they are not only factorizable but also proportional to the tree level yukawa couplings .
the second class is that of factorizable but not proportional to yukawa couplings contributions , these are for instance the ones in which there is a non - zero area in just one sub - torus .
finally there are other contributions that do not factories , corresponding for instance to non - zero area of the blue type in fig .
[ crossvsnoncross ] in more than one torus .
the very rich structure present in four - point amplitudes propagates through quantum corrections to the yukawa couplings . even though a full string calculation is possible
, we will treat threshold corrections at the field theory level , plugging the values of the chirality changing amplitudes we have just computed as effective vertices .
the one loop corrected yukawa couplings have then a flavor dependence shown in fig .
[ oneloopyuk ] .
corrections proportional to the yukawa couplings give no new structure but just renormalizes the tree level values , factorizable but non proportional to yukawa corrections increase the rank of the yukawa matrix in one unit , therefore giving mass to the second generation .
finally , non - factorizable corrections give masses to all the three generations allowing therefore for a fully realistic spectrum .
in fact , the different contributions we have just mentioned , give a rationale for the hierarchical spectrum observed in nature .
masses for the third family are generated at tree level , thus it is natural for them to be large .
the first two generations , getting masses at one loop , are naturally much lighter than the third one .
furthermore , an extra hierarchy can naturally arise if the non - factorizable corrections are suppressed with respect to the factorizable ones . in order to test in a quantitative way our assertions ,
we have performed a fit to the quark masses and mixing angles reproducing reasonably well the observed spectrum and obtaining in this way a well defined flavor pattern that will allow us to give unambiguous predictions for the phenomenological implications of models with intersecting branes .
the relevant ( for flavor ) parameters defining our model are , the horizontal and vertical ratios in the second and third torus , the position of the right brane in the second , @xmath34 and the third torus , @xmath35 , the position of the left brane in the third torus , @xmath36 , and the ratio of the up - type to down - type higgs vevs , @xmath37 . a reasonable value of the quark masses and mixing angles , @xmath38 where we have included a global normalization factor 0.95 in the yukawa couplings , is obtained for the following parameters ( dimensional parameters are expressed in string units ) @xmath39 in this particular example ,
the full flavor structure is fixed , all the values of the relevant moduli , yukawa couplings , rotation matrices , etc .
are known in terms of the string scale .
we can therefore compute without any ambiguity the values of any observable as a function of the string scale and obtain in that way precise bounds on it .
we would like to make one further remark before discussing the bounds on the string scale in these models .
as we have said , the vast majority of the realistic models with intersecting branes presented in the literature so far are non - supersymmetric .
it is to those that the specific mechanism for non - trivial fermion mass generation through yukawa threshold effects apply .
one should however consider the possibility of supersymmetric realistic models ( as we will see in the next section the very constraining bounds on the string scale in these models makes it highly desirable for them to be supersymmetric ) in which case non - renormalization theorems prevent threshold effects in yukawa couplings .
the very model we have presented possess a supersymmetric spectrum if the ratio of horizontal to vertical radii is equal in the second and third tori ( hidden branes that break all the supersymmetries are anyway present in this model so even though the contribution we have computed would vanish in the `` supersymmetric configuration '' , there would be similar ones from strings stretching between the hidden and the mssm sectors )
. nevertheless , even in fully supersymmetric configurations , there are still sources of non - trivial flavor structure that can potentially lead to a realistic fermion spectrum .
one possible source is for instance soft breaking a - terms .
they have a different flavor structure than yukawa couplings and this could be propagated through supersymmetric quantum loops to the yukawa couplings .
a detailed study of this and other effects is still necessary to see if realistic flavor patterns can be obtained as well in supersymmetric models .
once a realistic pattern of fermion masses and mixing angles has been developed we can compute , in a well - defined way , the effects of four - point string amplitudes in different observables in order to establish lower bounds on the string scale .
the presence of flavor changing processes makes it more likely that rare processes are the most constraining ones .
nevertheless we will also consider the contribution to flavor conserving processes that give independent yet quite strict bounds on the string scale .
we start our discussion with the former .
tree level flavor violating four - point amplitudes directly contribute to rare processes like meson oscillations and rare lepton and meson decays .
these processes are loop and sometimes gim suppressed in the standard model and therefore very small .
they are also extremely well constrained experimentally what makes of them a very powerful probe of flavor violating physics beyond the standard model . for this analysis , we have closely followed the discussion in ref .
@xcite , where the phenomenological implications of an extra @xmath40 gauge boson with family non - universal couplings was considered .
of course , it has been extended to take into account not only the effects of gauge boson kk modes but also those of string instantons . instead of giving a full account of the effect in each of the observables that have been considered
, we will discuss in some detail the case of meson oscillations and refer to @xcite for a more in - depth discussion .
the mass splitting for a meson with quark content @xmath41 , in the vacuum insertion approximation , reads , @xmath42 - \left[\frac{1}{2}+\frac{1}{3 } \left ( \frac{m_p}{m_{q_i}+m_{q_j } } \right)^2 \right ] \mathrm{re } a^{lr}_{ijij } \right\},\ ] ] where @xmath43 and @xmath44 are , respectively the mass and decay constant of the meson . here
@xmath45 are the dimensionless coefficients parameterizing the four fermion operators ( with @xmath46 factored out )
. indirect cp violation in the kaon system , which has been measured as well with extreme accuracy , is parametrized by @xmath47 - \left[\frac{1}{2}+\frac{1}{3 } \left ( \frac{m_k}{m_{d}+m_{s } } \right)^2 \right ] \mathrm{im } a^{lr}_{dsds } \right\}.\ ] ] we compute the limits on the string scale requiring that the new contribution is smaller that the experimental value .
this implies the following restrictions on the different coefficients , * kaon mass splitting @xmath48 - 17.1 \mathrm{re } \left [ a^{lr}_{dsds } \right ] | \lesssim 3.3 \times 10^{-7 } \mbox { tev}^{-2},\ ] ] * @xmath16 mass splitting @xmath49 -3\ , \mathrm{re}\big [ a^{lr}_{dbdb } \big ] \right| \lesssim 2 \times 10^{-6 } \mbox { tev}^{-2},\ ] ] * @xmath50 mass splitting @xmath51 -3\ , \mathrm{re}\big [ a^{lr}_{sbsb } \big ] \right| \lesssim 6.6 \times 10^{-5 } \mbox { tev}^{-2},\ ] ] * @xmath52 mass splitting @xmath53 -3.9\ , \mathrm{re}\big [ a^{lr}_{ucuc } \big ] \right| \lesssim 3.3 \times 10^{-6 } \mbox { tev}^{-2},\ ] ] * kaon cp violation @xmath54 -17.1\ , \mathrm{im}\big [ a^{lr}_{dsds } \big ] \right| \lesssim 2.6 \times 10^{-9 } \mbox { tev}^{-2}.\ ] ] it is clear that if the coefficients of the 4 fermion operators are order one ( times @xmath55 ) , the kaon system puts a constraint @xmath56 tev . using the specific values of the rotation matrices found before we show in table [ boundsfcnc : table ] that these estimations are indeed right .
the list observables is completed with @xmath57 coherent conversion in atoms , tau decays , and leptonic and semi - leptonic meson decays .
having developed a theory of fermion masses and mixing angles for the quark sector , we can only make an estimation of the bounds implied by these leptonic observables . nevertheless , the order of magnitude of the bounds on the string scale implied by those is similar of the one found for quark observables , as shown in table [ boundsfcnc : table ] .
flavor violating observables have proved extremely constraining in models with intersecting branes .
the bounds obtained in the previous section should be considered as typical bounds for these models .
the long list of observables we have studied makes it highly implausible that a choice of parameters ( moduli ) can be made for which _ all _ the flavor violating observables are strongly suppressed . in any case , in order to try and avoid that situation and for the sake of completeness we consider now the effects on flavor preserving observables .
these effects rely more on very generic features of models with intersecting branes than in the particular structure of flavor .
these features are , among others , the presence of several @xmath40 gauge fields and the appearance of right handed neutrinos .
for that reason we shall not use the detailed form of the rotation matrices to perform a very precise computation but will just estimate the effects and the corresponding bounds on the string scale .
we will start with a flavor conserving but cp violating observable , electric dipole moment ( edm ) of ( mainly mercury ) atoms .
edm measurements mostly constrain operators of the type @xmath58 , @xmath59 , etc .
let us concentrate on the first transition for definiteness .
the relevant lagrangian is therefore @xmath60 using the recent analysis in ref .
@xcite we can put the following bound on the corresponding coefficient , @xmath61 from the mercury edm .
the bounds on the other coefficients and the one coming from the neutron edm are weaker than this one . including typical values for the mixing angles we obtain the following ( conservative ) estimation for the bound on the string scale @xmath62 the next effect we are going to consider is supernova cooling .
for that we will use two rather generic features of models with intersecting branes , the presence of right handed neutrinos and the appearance of new @xmath40 gauge groups that get masses ( and therefore mix among themselves ) through a generalized green - schwarz mechanism .
the emission of right - handed neutrinos during supernova collapse can affect the rate of cooling .
this implies the following constraint , for one species of light dirac neutrinos , @xmath63 where @xmath64 is the scale of the following quark - neutrino interaction @xmath65 this effective interaction can be mediated by the `` right '' gauge boson ( the one corresponding to the @xmath40 gauge group in the right brane ) at tree level as well as through the one loop mixing between the @xmath66 and the `` leptonic '' gauge bosons as represented in fig .
[ supernova ] .
the details of course depend on the particular values of the masses and couplings of the corresponding gauge bosons to the fermions .
these were discussed in @xcite .
a good estimation is however possible to make , taking into account the different possibilities .
we obtain the following ( again conservative ) bound from the tree level process @xmath67 whereas the loop - suppressed process gives @xmath68 other processes like lep constraints on contact interactions or the effects of the extra @xmath40 s on the @xmath69 parameter , lead to less stringent bounds on the string scale of the order of the tev .
models with intersecting branes have many phenomenologically appealing features . in order to further test their potential for fully realistic models , we have considered their flavor structure . in this way we have been able to compute in an unambiguous way their contributions to a number of flavor violating and preserving operators .
the resulting bounds are very stringent @xmath70 this feature will be generic in any string model in which the different generations of fermions live at separate points in the compact space . in that case , unless flavor violations are strongly suppressed ( an example of this situation occurs in the warped case ) for any reason , the high string scale implied by them is difficult to make compatible with a stabilized electroweak scale in non - supersymmetric models .
we should not be disappointed by these results .
they should instead encourage us in the search of realistic supersymmetric models for which much higher string and compactification scales are stable against radiative corrections .
it is a great pleasure to thank s.a .
abel , o. lebedev and m. masip for collaboration and i. navarro for useful discussions .
this work has been funded by pparc .
99 r. blumenhagen , l. goerlich , b. kors and d. lust , jhep * 0010 * ( 2000 ) 006 [ arxiv : hep - th/0007024 ] ; s. forste , g. honecker and r. schreyer , nucl .
b * 593 * ( 2001 ) 127 [ arxiv : hep - th/0008250 ] ; g. aldazabal , s. franco , l. e. ibaez , r. rabadn and a. m. uranga , jhep * 0102 * ( 2001 ) 047 [ arxiv : hep - ph/0011132 ] ; g. aldazabal , s. franco , l. e. ibaez , r. rabadn and a. m. uranga , j. math .
* 42 * ( 2001 ) 3103 [ arxiv : hep - th/0011073 ] . s. abel and j. santiago , j. phys .
g * 30 * ( 2004 ) r83 [ arxiv : hep - ph/0404237 ] .
m. cvetic , g. shiu and a. m. uranga , phys . rev
. lett .
* 87 * ( 2001 ) 201801 [ arxiv : hep - th/0107143 ] ; m. cvetic , g. shiu and a. m. uranga , nucl . phys .
b * 615 * ( 2001 ) 3 [ arxiv : hep - th/0107166 ] ; d. cremades , l. e. ibaez and f. marchesano , jhep * 0207 * ( 2002 ) 009 [ arxiv : hep - th/0201205 ] ; r. blumenhagen , l. gorlich and t. ott , jhep * 0301 * ( 2003 ) 021 [ arxiv : hep - th/0211059 ] ; m. cvetic , i. papadimitriou and g. shiu , arxiv : hep - th/0212177 ; m. cvetic and i. papadimitriou , phys . rev .
d * 67 * ( 2003 ) 126006 [ arxiv : hep - th/0303197 ] ; m. cvetic , p. langacker and j. wang , phys .
d * 68 * ( 2003 ) 046002 [ arxiv : hep - th/0303208 ] ; k. behrndt and m. cvetic , nucl .
b * 676 * ( 2004 ) 149 [ arxiv : hep - th/0308045 ] .
g. honecker , nucl .
b * 666 * ( 2003 ) 175 [ arxiv : hep - th/0303015 ] .
d. cremades , l. e. ibaez and f. marchesano , jhep * 0307 * ( 2003 ) 038 [ arxiv : hep - th/0302105 ] .
s. a. abel , m. masip and j. santiago , jhep * 0304 * ( 2003 ) 057 [ arxiv : hep - ph/0303087 ] .
m. cvetic and i. papadimitriou , phys .
d * 68 * ( 2003 ) 046001 [ arxiv : hep - th/0303083 ] . s. a. abel and a. w. owen , nucl . phys .
b * 663 * ( 2003 ) 197 [ arxiv : hep - th/0303124 ] ; s. a. abel and a. w. owen , arxiv : hep - th/0310257 . | we review the very stringent lower bounds on the string scale that arise from flavor considerations in models with intersecting branes . despite the absence of a realistic flavor structure at tree level ,
flavor changing interactions induce a non - trivial pattern of fermion masses and mixing angles when quantum corrections are taken into account .
the resulting realistic theory of flavor allows us to constrain , in an unambiguous way , the string scale up to levels difficult to reconcile non - supersymmetric models . |
identification of sources of the ultra - high energy cosmic rays ( uhecr ) is a key to understanding their nature .
the observed small - scale clustering of uhecr @xcite suggests that already existing data @xcite may contain information sufficient to make first steps in this direction . in a series of papers @xcite
we have shown that there exist significant correlations between the arrival directions of uhecr and bl lacertae objects ( bl lacs ) , and therefore bl lacs are likely to be sources of uhecr .
although present small dataset does not allow to determine with certainty the details of uhecr production and propagation , interesting hints may be obtained @xcite .
the first evidence of connection between uhecr and bl lacs was found in ref .
@xcite where we have shown that the combined set of agasa events with @xmath0 ev and yakutsk events with @xmath1 ev ( identified previously as the sets with largest clustering ) correlates strongly with most powerful bl lacs .
the statistical significance of this correlation was found to be @xmath2 with penalty factor included ( for explanation of penalty factor see below ) . in the recent paper by evans ,
ferrer and sarkar ( efs ) @xcite the validity of this result was called into question .
the rational part of the criticism of efs boils down to the following two issues : i ) choice of cuts on bl lacs and/or calculation of the penalty factor ii ) choice of the cosmic ray set .
below we address these issues and demonstrate that criticism of efs is unjustified .
we also clarify some frequently arising questions related to our analysis .
in statistical analysis , one nearly always has to make cuts in order to improve signal - to - noise ratio .
the question is how to take them into account correctly .
a common wisdom is that cuts should be fixed _ a priori _ , i.e. , based on theoretical considerations . in that case the cuts simply limit the number of data points , but do not alter the calculation of probabilities
. not always this is possible .
for instance , in the case of uhecrs , their acceleration mechanism is not known .
how to impose cuts in a catalogue of astrophysical objects in order to select actual uhecr emitters ? in which band radio , optical , x - ray , @xmath3 , tev ? to which flux limit ? in ref .
@xcite we proposed an approach to this problem which consists in adjusting cut(s ) so as _ to maximize _ the signal , and then compensating this cut adjustment by a _
penalty factor_. a penalty factor takes into account the `` number of independent attempts '' made when searching for the best signal .
for instance , if two independent catalogues were tried , the penalty factor is 2 ( in the limit of small probabilities ) .
if @xmath4 catalogues tried are not independent , as in the case of cuts in a single catalogue , the penalty factor is smaller than @xmath4 and should be calculated by means of the monte - carlo simulation . in correlation analysis of cosmic rays ,
the quantity of interest is the probability @xmath5 that the observed excess of cosmic rays around source positions is the result of a chance coincidence .
this probability depends on cuts made in the source catalogue ; the latter are adjusted so that the probability is minimum ( correlations are maximum ) .
the cut adjustment should be compensated by the penalty factor which is calculated as follows .
a random set of cosmic rays is generated and treated exactly as the real data : the same cuts are tried and the resulting minimum probability @xmath6 is determined .
this procedure is then repeated for a large number of random `` cosmic ray '' sets , and @xmath7 is determined each time . the number of occurrences of a value @xmath8 is then counted as a function of @xmath9 . divided by the total number of sets , this gives the probability @xmath10 that the adjustments of cuts produces @xmath11 for a random set of cosmic rays .
@xmath12 is the correct measure of the significance of observed correlations .
when no adjustment of cuts is made one obviously has @xmath13 , and the quantitative measure of correlation is the probability @xmath9 itself . when cuts are adjusted , small probabilities appear more often by construction , and this relation is modified .
the modification can be written in terms of the penalty factor @xmath14 , @xmath15 in the presence of cut adjustment , the significance of correlations is determined by the product @xmath16 . to summarize
, all parameters characterizing uhecr sources and uhecrs themselves can be subject to cuts .
some of these cuts can be decided a priori ; they reflect ( and depend on ) physical assumptions made .
these cuts do not require penalty .
one can call these cuts _
fixed_. alternatively , in the absence of physical arguments , the cut should be chosen ( adjusted ) so as to maximize the signal .
these cuts can be called _
adjustable_. these are the adjustable cuts which imply non - trivial penalty factors ; the latter can be calculated in a way explained above . in our analysis @xcite ,
the cosmic ray set was fixed on physical grounds using results of previous publications @xcite , and was not adjusted in search for correlations .
the motivation was as follows .
it was observed earlier @xcite that highest - energy cosmic rays exhibit remarkable property : they auto - correlate on the angular scale consistent with the angular resolution of detectors .
the mere existence of these correlations suggests that a relatively small number of point sources of cosmic rays is contributing a considerable fraction of the uhecr flux @xcite , and that the trace of the original directions to sources is not lost completely .
the identification of these sources via cross - correlation analysis may therefore be possible . within the hypothesis that uhecr clusters are due to point sources ,
the strongest correlation signal is expected with the uhecr set which has most significant auto - correlations .
the latter requirement selects @xcite agasa events with @xmath0 ev and yakutsk events with @xmath17 ev . the observed angular size of clusters also fixes the angular scale at which correlations with sources are expected .
the fact that agasa and yakutsk sets have different energy cuts in this approach is , of course , disturbing .
it may , however , be explained by different energy calibration of the two experiments , or smaller number of the yakutsk events . in either case , within our assumptions it is inconsistent to change energy cuts or to drop yakutsk events once autocorrelations are found in this set .
the authors of ref .
@xcite have rejected the yakutsk events because yakutsk array has angular resolution not as good as agasa . by itself , worse angular resolution does not imply that correlations with sources _ must _ be absent in yakutsk set : even though the angular resolution is worse , the density of uhecr events around actual sources is larger as compared to a random set , and one has an excess in counts even at small angles .
the actual presence of autocorrelations in yakutsk data supports this statement .
therefore , it is not correct that `` the correlations found at smaller @xmath18 angles can not be meaningful '' @xcite .
the authors of ref .
@xcite have essentially reproduced our monte - carlo simulations : solid curve of fig . 3 of ref .
@xcite agrees with our calculations if yakutsk events are discarded .
they concluded , however , that we have underestimated the penalty factor .
this conclusion was based on i ) a comparison of correlations in the case when one particular set of cuts was chosen to the case when no cuts on bl lacs were made at all ii ) smaller correlations of bl lacs with a different set of uhecr .
neither such a comparison , nor the dependence on the uhecr set have anything to do with the penalty factor .
the uhecr set was _ fixed _ as described above and was not adjusted during the calculation , thus requiring no penalty . the penalty associated with the adjustment of bl lac set
should be _ calculated _ as described above .
this calculation was performed in our paper @xcite as follows .
the most complete catalog of qso @xcite contains 306 confirmed bl lacs for which the apparent magnitude , redshift ( where known ) and 6 cm radio flux are listed ( for several objects the 11 cm radio flux is also given ) . in ref .
@xcite we considered bl lacs with @xmath19 as suggested by statistics of clustering .
( we also examined a complimetary small set without further cut adjustment .
this `` independent attempt '' adds one to the penalty factor . )
we adjusted cuts on magnitude and 6 cm radio flux in the following way .
first , the grid of @xmath20 cuts was fixed .
namely , the cut on magnitude was varied from @xmath21 to @xmath22 with the step @xmath23 ; simultaneously the cut on 6 cm radio flux was changed from 0.21 jy to 0.01 jy with the step 0.02 jy .
note that further `` refining '' of this set of cuts would not change the results as the sets of bl lacs obtained would overlap almost completely .
the lowest probability ( best correlation ) found @xcite in the real data in this way equals to @xmath24 .. this is merely misunderstanding of our paper . ] to calculate the penalty factor associated with cut adjustment , a random configuration of 65 uhecr events was generated .
then its correlation with all @xmath20 subsets of bl lacs was examined and the minimum probability @xmath25 over @xmath20 cases was found and recorded .
the whole procedure was repeated for the next random configuration .
a total of @xmath26 random configurations were treated in this way .
the number of occurrences @xmath27 of @xmath28 was calculated as a function of @xmath9 .
divided by @xmath29 it determines the corrected probability @xmath10 and the penalty factor @xmath14 as defined in eq .
( [ eq : penaltydef ] ) .
the resulting penalty factor @xmath14 is shown in fig . [
fig : penalty2 ] .
( 240,170)(0,0 ) the error bars correspond to statistical errors in determination of @xmath27 . taking the penalty factor at the minimum probability obtained for the real data , @xmath30 and adding 1 as penalty for bl lacs with @xmath31
as explained above , one finds the significance of correlations including penalty , @xmath32 the significance is @xmath33 in terms of gaussian distribution .
it should be pointed out that , in order to get the correct significance , the penalty factor should be multiplied by the _ minimum _
probability obtained for the real data with the same set of cuts .
we stress again that the absence of correlations of cosmic rays with the entire sample of known bl lacs does not mean anything .
physically , cosmic rays correlate with the brightest bl lacs . including dimmer bl lacs in the analysis
dilutes the correlations , and this is the only conclusion one draws from the analysis made in ref .
@xcite .
the authors of ref .
@xcite have chosen grbs as a control set where correlations are not expected ; it has to be compared with the case of bl lacs . from the fact that correlations with the _ whole _ catalogue are absent in both cases while there are many coincidences , the authors of ref .
@xcite concluded that grbs correlate with uhecr `` just as well as do bl lacs ! '' , implying that correlations with bl lacs are due to cut adjustment .
this conclusion can be tested by monte - carlo simulation : one has to check whether cut adjustment leads to apparent correlations with grbs .
it does not , as we now demonstrate .
the fourth batse gamma - ray burst catalog @xcite contains 1637 objects from which we cut out , in sequential order ( according to observation date and time ) , 5 non - overlapping sets of 306 objects each ( recall that 306 is the number of confirmed bl lacs in the catalogue @xcite ) . in each set
we have looked for maximum correlations by optimizing cuts .
we present two different tests : one - dimensional cut which allows transparent graphical representation , and two - dimensional cut which mimics more precisely the case of bl lacs .
we took the same maximally auto - correlated uhecr data set as in refs .
@xcite . in the first test
we replace the @xmath20 cuts on magnitude and radio flux by 121 cut on the total number of objects ( we take them in sequential order , so that first few cuts correspond to including 3 , 6 , 9 ... first objects , and the last cut corresponds to including all 306 objects ) . like in the case of bl lacs
, we then calculate the probability of the observed excess of uhecrs around grbs as a function of the cut .
the results are presented in fig .
[ fig : grb ] .
the smallest probability found in all 5 sets is @xmath34 ( which by itself is not statistically significant ) . in view of the similar number of cuts in two cases , fig . 1 provides an estimate for the penalty factor , @xmath35 , in a single set of 306 grbs .
multiplying the best probability , the penalty factor and the number of independent grb sets one gets a number of order 1 .
( 210,200)(10,0 ) ( 0,0 ) to mimic the case of bl lacs we perform a similar test with _ two _ independent cuts .
the first cut is done as before on the sequential number of the event .
the second cut is imposed on the time ( in seconds of the day , ut ) of the event .
note that these two cuts are virtually uncorrelated since batse registered roughly one event per day .
we have searched for the best correlation signal on the grid of @xmath20 equidistant cuts from 1 to 306 and from 1 to 86400 , respectively . in the above five catalogs consisting of 306 grbs each ,
the following lowest probabilities were found : 9.5% , 1.4% , 20% , 1.8% , 24% .
multiplying the lowest of these probabilities 0.014 by the penalty factor @xmath36 and by the number of independent grb sets , we again obtain a number of order 1 .
these examples are a good illustration of compensation between the effects due to cut adjustment and the penalty factor .
when compensated by the penalty factor , cut selection does not introduce apparent correlations when in reality they are absent . to avoid confusion , we note again that we used grbs here as a control set , and imposed physically unmotivated cuts on purpose .
in this section we discuss some subtleties of correlation analysis which were not mentioned in ref .
we hope that this will answer a number frequently arising questions regarding our work and correlation analysis in general .
a catalogue of astrophysical objects is called complete when it is believed to contain more than , say , 90% of all existing objects of a given type , in a given region of the sky , and down to some fixed level of luminosity in a given waveband(s ) .
the catalogue @xcite , used in our analysis , is not complete simply because it contains all objects known up to date without further selection @xcite .
it is a common wisdom that completeness of the catalogue is crucial for statistical analysis .
the reason is obvious : in an incomplete catalogue , the distribution of objects in brightness , position , etc .
, reflects not only their actual abundance , but also observational bias .
when one studies , for instance , the evolution of abundance of some objects with time , one has to make sure that logn / logs dependence reflects actual change in spatial density rather than the fact that remote ( and therefore dim ) objects are more difficult to observe . indeed ,
as it is , the catalogue @xcite is not suitable for any analysis where _ statistical properties of bl lacs _ are of interest . in the correlation analysis of uhecr
the question is different : given a set of candidate sources ( bl lacs in the case at hand ) , is the distribution of cosmic rays random , or does it peak around bl lac positions ? clearly , these are the _ statistical properties of cosmic rays _ which are of interest in that case , not the statistical properties of bl lacs . technically speaking , when calculating the correlation function by the algorithm of ref .
@xcite , the cosmic ray directions are simulated , while source positions are held fixed .
for this reason the method is applicable without change to studying , for instance , a correlation of uhecr with one particular direction ( say , the galactic center ) , i.e. in the case when the notion of completeness does not apply .
completeness is not necessary for _ establishing the fact _ of correlations : if statistically significant correlations are found with an incomplete catalogue , this is a real signal .
incompleteness of the catalogue of bl lacs can not be a source of correlations with uhecr .
indeed , the objects which have to be added ( removed ) from a catalogue to make it complete are absent from ( present in ) the catalogue for reasons not related to cosmic rays .
therefore , correlations with uhecr can only be weakened by such an incompleteness . the uhecr set used in our calculations
is known to contain event clusters in fact , it was chosen in such a way that auto - correlations are maximum .
when studying cross - correlations of such a set with potential sources , one has to be careful to take clusters into account when calculating the probability of chance coincidence .
namely , one has to make sure that the signal observed in the data is not due to chance coincidence of _ clusters _ of cosmic ray events with candidate sources . for a given cluster ,
such a coincidence is roughly as probable as for a given single event , but contributes more into correlation function .
this could produce artificial enhancement of correlations if not taken into account in the monte - carlo simulation . in our calculations @xcite this problem
is solved by introducing in each monte - carlo cosmic ray set the same number of doublets and triplets as there are in the real data
. then _ chance _ coincidences between clusters and candidate sources happen in simulated sets as often as in the real data , and thus are correctly accounted for .
this is confirmed by calculations presented in sect .
[ sect : grb ] where correlations with grb catalogs do not appear despite auto - correlations in the uhecr dataset .
the observed angular size of clusters suggests the angular scale at which correlations with sources are expected . in our analysis
@xcite we have fixed it to the previously published @xcite agasa value of @xmath37 .
this value is treated as not adjustable .
when the angular resolution of the experiment is known , there exists a preferred choice of the angular scale the scale at which correlations are _ expected _ to be maximum .
this angle can be determined by means of the monte - carlo simulation as follows .
one has to generate cosmic ray sets which are correlated with sources ( taking into account the experimental angular resolution ) and then measure angle at which the correlation signal is maximum . in the case of agasa events
this procedure gives the result very close to @xmath37 ( see dotted curve on fig .
3 of ref .
@xcite ) , so in fact the choice of ref .
@xcite was close to optimal .
alternatively , one may choose not to fix the angular scale and treat it as a free parameter .
then one should adjust it to maximize the correlation and calculate corresponding penalty factor .
a serious problem of this approach is that the result would depend on the limits within which the angular scale is varied .
so finally one would have to input in one or the other way the information about expected angular scale of correlations in order to obtain a definite answer .
we do not follow this approach in our calculations .
we believe that we have accounted for all contributions to the penalty factor ( `` effective number of tries '' ) _ directly related _ to our work . still , the obtained significance may be somewhat overestimated .
the point is that we are not the first who are looking for correlations between uhecr and astrophysical objects .
some of these attempts have been published in the literature ; the others may have never been reported .
all these attempts should contribute , in principle , into the penalty factor .
however , it does not seem possible to account correctly for all such contributions .
the way around this problem is obvious .
first , only very low values of @xmath38 should be interpreted as a signal ( in our calculations we considered @xmath39 to be sufficiently low to report our results ) .
second , and more important , the results have to be confirmed with a new independent data set .
the correlations found in ref .
@xcite for a particular set of uhecr and bl lacs suggest , strictly speaking , only that uhecr and bl lacs are connected to each other .
they say little about the acceleration mechanism , particle nature and other relevant physical parameters .
these questions can be fully addressed only with much larger dataset than available now .
some attempts , however , can be made .
the final purpose is to arrive at the end at a consistent picture which incorporates all known features of uhecr .
here is a sketch of these attempts , each bringing additional evidence of the connection between uhecr and bl lacs . in ref .
@xcite it was noted that cuts in bl lac catalogue , chosen in ref .
@xcite so as to maximize correlations with uhecr , select automatically @xmath3-ray loud bl lacs .
when this observation is consistently elaborated , i.e. a sub - sample of the catalogue @xcite is selected on the basis of a single criterium , namely , the cross - correlation with the egret sources , the resulting subset of 14 @xmath3-ray loud bl lacs correlates with uhecr at the level of @xmath40 of chance coincidence .
this number can not , of course , be interpreted as the significance of correlations between uhecr and bl lacs because of _ a posteriori _ selection ; rather the conclusion is that @xmath3-ray loudness may be a distinctive feature of those bl lacs which are uhecr accelerators . in ref .
@xcite an attempt was made to determine the charge composition of uhecr by reconstructing actual arrival directions of uhecr particles bent in the galactic magnetic field .
the idea was that such a reconstruction should improve correlations of uhecr with bl lacs if the latter are the sources .
substantial improvement was indeed observed for particles with the charge + 1 , which is an indication of the presence of protons . in ref .
@xcite it was observed that if correlations of uhecr and bl lacs were due to chance coincidence , the coinciding rays would be distributed over the sky randomly , reflecting only the local density of bl lacs and exposure of a cosmic ray experiment .
thus , any significant deviation in the distribution of _ correlating _ rays over the sky from this expectation speaks in favor of real physical connection between cosmic rays and bl lacs .
in fact , the uhecrs correlating with bl lacs form two `` spots '' , with low probability to occur by chance @xcite .
this non - uniformity of the distribution of correlating rays may be due to several factors : ( 1 ) anisotropy of extragalactic magnetic fields at scales of order @xmath41 mpc ; ( 2 ) poor knowledge of the galactic magnetic field in some areas of the sky ; ( 3 ) fluctuations in the space distribution of the nearest sources .
according to textbooks , any successful statistical analysis consists in formulation of a `` null hypothesis '' and its subsequent _ falsification _ , at some confidence level , by comparing to the experimental data . in the case at hand
the `` null hypothesis '' which is being tested is that bl lacs ( and , in particular , any subset of them ) and uhecr are uncorrelated .
it takes _ only one _ counter - example to disprove a hypothesis , while `` pro - examples '' _ do not _ prove its validity .
an illustration of this general rule has been discussed in sect .
ii : the absence of significant correlations of uhecr with the whole bl lac catalogue _ does not prove _ that uhecr and bl lacs are uncorrelated ( i.e. , that there is no subset of bl lacs which are sources of uhecr and thus correlate with them ) .
having found such counter - example ( i.e. the case when correlation is significant ) , the only thing which can be concluded , to a certain confidence level , is that uhecr and a particular subset of bl lacs are correlated .
the nature and physical implications of these correlations have to be studied separately by formulating and testing different `` null hypothesis '' .
refs.@xcite are first attempts in this direction .
= 400 we are grateful to v. rubakov , d. semikoz , m. shaposhnikov and p. veron for reading the manuscript and valuable comments . the work of p.t . is supported in part by the swiss science foundation , grant 20 - 67958.02 .
the authors of ref .
@xcite have misinterpreted our earlier paper @xcite which concerns the statistics of clustering of uhecr and the number of their sources . in view of the growing confusion we would like to clarify this issue . in ref .
@xcite one reads : `` @xmath42observed occurrence of clusters of uhecr@xmath43 was used to estimate the spatial density of sources to be @xmath44 @xcite .
this would obviously place stringent constraints on candidate astrophysical sources , e.g. @xmath3-ray bursts ( grbs ) have a spatial density of only @xmath45 .
however , a more careful analysis @xcite shows that the uncertainties in this estimate are very large .
the true number is @xmath46 at the 68% ( 95% ) c.l .
; moreover relaxing the assumptions made , viz .
that the sources all have the same luminosity and a spectrum @xmath47 , increases the allowed range even further , e.g. to @xmath48 '' .
the actual situation is different from what is described in this extract .
first , intrinsic inaccuracy of the estimate was fully acknowledged in ref .
its physical reason is obvious : clustering is not sensitive to the number of dim sources .
the estimates of ref .
@xcite cited above are a good illustration of the point : _ upper _ limits are huge and strongly model - dependent .
second , the main point of ref .
@xcite was to show that there exists a _
model - independent lower bound _ on the number of sources .
this bound is presented in table 1 of ref .
@xcite : at 90% and 99% confidence levels the number of sources is @xmath49 and @xmath50 , respectively .
this is consistent with the calculations of ref .
@xcite performed in particular models .
99 w. evans , f. ferrer and s. sarkar , astro - ph/0212533 .
n. hayashida _ et al .
_ , phys .
* 77 * , 1000 ( 1996 ) .
m. takeda _ et al .
_ , astrophys .
j. * 522 * , 225 ( 1999 ) [ astro - ph/9902239 ] .
p. g. tinyakov and i. i. tkachev , jetp lett .
* 74 * , 1 ( 2001 ) [ pisma zh .
fiz . * 74 * , 3 ( 2001 ) ] , [ astro - ph/0102101 ] .
m. takeda et .
al . , proc . of 27th international cosmic ray conference ( hamburg , germany , 7 - 15 aug 2001 ) p. 341 .
m. takeda _ et al .
_ , phys .
lett . * 81 * , 1163 ( 1998 ) ; m.a .
lawrence , r.j.o .
reid and a.a .
watson , j. phys .
g : nucl .
phys . , * 17 * , 733 ( 1991 ) ; b.n
. afanasiev _ et al .
int . symp . on extremely high energy cosmic rays : astrophysics and future observatories , ed . by nagano ,
p.32 ( 1996 ) .
n. hayashida _ et al .
_ , astrophys .
j. * 522 * , 225 ( 1999 ) ( appendix ) [ astro - ph/0008102 ] .
p. g. tinyakov and i. i. tkachev , jetp lett .
* 74 * , 445 ( 2001 ) [ pisma zh
. eksp .
fiz . * 74 * , 499 ( 2001 ) ] [ astro - ph/0102476 ] .
p. tinyakov and i. tkachev , compact sources of uhecr , proc .
of 27th international cosmic ray conference ( hamburg , germany , 7 - 15 aug 2001 ) p. 547
; [ http://www.copernicus.org/icrc/he1.05.oral.htm ] .
p. g. tinyakov and i. i. tkachev , talk at taup 2001 : topics in astroparticle and underground physics , assergi , italy , 8 - 12 sep 2001 . published in : nucl .
suppl . * 110 * , 501 ( 2002 ) .
p. g. tinyakov and i. i. tkachev , astropart .
phys . * 18 * , 165 ( 2002 ) [ astro - ph/0111305 ] .
d. s. gorbunov , p. g. tinyakov , i. i. tkachev and s. v. troitsky , astrophys .
j. * 577 * , l93 ( 2002 ) [ astro - ph/0204360 ] .
tinyakov and i. tkachev , talk at international workshop on extremely high energy cosmic rays , wako , japan , 5 - 6 nov 2002 ; hep - ph/0212223 .
tinyakov and i. tkachev , talk at icrc 2003 , tsukuba , japan , 31 jul - 7 aug 2003 ; astro - ph/0305363 .
s. l. dubovsky , p. g. tinyakov and i. i. tkachev , phys .
lett . * 85 * , 1154 ( 2000 ) .
veron - cetty and p. veron , quasars and active galactic nuclei ( 9th ed . ) , eso scientific report ( 2000 ) .
[ http://vizier.u-strasbg.fr/viz-bin/cat?vii/215 ] w. s. paciesas _ et al .
_ , astro - ph/9903205 .
this catalogue is regularly updated , current edition is : m.p .
veron - cetty and p. veron , quasars and active galactic nuclei ( 10th ed . ) , astron .
* 374 * , 92 ( 2001 ) .
[ http://vizier.u-strasbg.fr/viz-bin/cat?vii/224 ] | in a series of papers we have found statistically significant correlations between arrival directions of ultra - high energy cosmic rays and bl lacertae objects .
recently , our calculations were partly repeated by evans , ferrer and sarkar @xcite with different conclusions .
we demonstrate that the criticism of ref .
@xcite is incorrect .
we also present the details of our method . |
the known population of qsos , thought to be the main source of ionizing photons throughout most of the history of the universe , falls short of maintaining the ionization of the intergalactic medium at redshifts beyond @xmath3 ( e.g. , @xcite ) , whereas the universe is known to be fully ionized back to at least @xmath4 ( e.g. , @xcite ) .
ionizing flux escaping from galaxies is often invoked to make up for the missing photons .
the highest redshift galaxy surveys are finding increasing numbers of @xmath5 galaxies .
however , to reach the number density of photons necessary for re - ionization may require either the existence of large numbers of ( still ) invisible sources ( e.g. , trenti , this conference ) , or rather large escape fractions for those types of galaxies currently observable ( e.g. , finkelstein , this conference ) .
the challenge posed by these conditions is exacerbated by a paucity of credible observations of any galaxies emitting a significant amount of ionizing radiation ( see , e.g. , @xcite for a recent discussion ) .
this may suggest that yet other types of sources may be responsible for reionization ( `` hidden '' agn , mini - halos ) , or that the galaxies at those lower redshifts currently accessible to observation for some reason are no longer leaking ionizing photons . with ordinary galaxies failing to ionize the universe
, it is instructive to briefly consider possible conditions that may facilitate the release of ionizing radiation .
these conditions generally fall into the following three categories : 1 .
enhanced production of ionizing photons , e.g. , by hotter , perhaps metal poor stars , or production of harder photons that experience a lower optical depth .
2 . creation of holes in the otherwise optically thick galactic disks or halos , through which ionizing radiation may escape , possibly through interactions that damage a gaseous halo , galactic winds , or through ionization by an agn .
3 . the formation in , or transport of the stellar sources of ionizing radiation into less optically thick environments .
examples may be any kind of intra - halo star formation , e.g. , in globular clusters or tidal tails . in the absence of the difficult , direct detections of leaking ionizing continuum
, the prime observational diagnostic of the above features would be the ly@xmath1 emission from neutral hydrogen
. a ly@xmath1 emission line from a gaseous halo that is leaking ionizing radiation may be spatially asymmetric , and possibly offset from the underlying sources of the ionizing continuum .
density and velocity disturbances of the halo gas , caused by galaxy interactions , could produce such asymmetries , as could multiple sources of ionizing radiation moving about in a joint gaseous halo .
the presence of the latter may also be evident from broad band detections of multiple stellar continuum objects , perhaps in the form of disturbed stellar populations , e.g. , tidal features , and young stellar populations , the formation of which may have been triggered by interactions .
probable detections of some of the above phenomena have recently come from ultra - deep spectroscopic blind searches for ly@xmath1 emission at redshift @xmath6 . in a reversal of the usual search strategy for ly@xmath1 emission ( narrow band imaging combined with spectroscopic followup )
, a spectroscopic blind survey detects ly@xmath1 emission line features in a randomly positioned , deep 2-dimensional long slit spectrum .
underlying continuum sources can be inspected in existing space - based imaging at the slit position .
because of its superior ability to suppress the sky background , this method is ideally suited for finding faint galaxies ( and has been able to uncover a large number density of @xmath2 ly@xmath1 emitters with luminosities down to @xmath7 ; @xcite ) . as a by - product
, this approach also uncovers very extended , low surface brightness ly@xmath1 emitters that would have escaped traditional searches . in a deep 61.4 h long slit spectrum , obtained with the magellan ldss3 instrument in the hudf
, we have discovered three such extended peculiar ly@xmath1 emitters in a volume of 2056 h@xmath8 mpc@xmath9 ( comoving ) .
the 2-d spectrum shows ly@xmath1 emission by these objects across several tens of kpc , with total fluxes on the order of a few @xmath10 erg @xmath11s@xmath12 .
while one of the objects turned out to be an obscured qso ( and will be described in a future publication ) , the other two which we shall describe here appear to be related to interacting galaxies , with no obvious signs of the presence of an agn . to illustrate the differences between
these extended , asymmetric emitters and standard lyman @xmath1 emitting galaxies we show in fig .
[ compares ] parts of the spectra of the two extended emitters discussed here in the panels on the left , and two other bright , previously known lyman break galaxies with ly@xmath1 emission on the right side .
the differences are striking , in that the lyman break galaxies show very compact , symmetric , spatial ly@xmath1 profiles , and spectral profiles with the usual red - dominated peak .
models of ly@xmath1 radiative transfer ( @xcite , @xcite ) show that this pattern is consistent with a single compact source of ionizing photons embedded in an hi halo highly optically thick to ionizing and ly@xmath1 photons .
in contrast , the extended emitters are spatially asymmetric , with a complex , multi - component spectral profile .
evidence to be described below suggests that the latter objects are gaseous halos where the velocity and density structure of the gas has been disturbed by interactions and multiple internal sources of ionizing and ly@xmath1 photons .
emission line regions of the two extended , asymmetric objects ( left column ) discussed here , and two more normal , compact ly@xmath1 emitters from galaxies that could have been detected by either lyman break or narrow band filter surveys ( right column ) . the latter , much more typical galaxies appear to have spatially symmetric halos , a more orderly velocity field , and can be explained by single sources of ionization . the galaxy in the bottom left panel is referred to below as `` case i '' , the one in the top left panel as `` case ii''.[compares ] ] emission line region of the 3.34 emitter ( left ) and a v band image from the goods - s survey ( @xcite ) of the underlying galaxy .
the emitter corresponds to a v=27 galaxy with stellar tails sticking out horizontally .
[ sys1 ] ] emission in the z=3.34 object relates to the escape of ionizing radiation : frontal view through the slit ( left ) , and view perpendicular to the line - of - sight ( center - right ) .
the compact ly@xmath1 emission line propagating from the center of the galaxy ( `` red core '' ) is partly absorbed by an optically thick damped ly@xmath1 region .
the fluorescent screen and filament represent external gaseous structures falling into the galaxy on the far side from the observer .
they are irradiated by ionizing photons escaping from a hole in the back of the gaseous halo and fluoresce in ly@xmath1 .
[ sys1_scheme ] ] the first case shows a `` fan '' of low surface brightness emission extending over about 35 kpc ( proper ; fig .
[ sys1 ] ) .
the underlying galaxy is a disturbed v=27 object with apparent stellar tails sticking out on two sides .
the relatively complex spectral and spatial shape of the ly@xmath1 emission can be interpreted as being caused by gas falling in from the back onto the galaxy ( fig .
[ sys1_scheme ] ) .
the in - falling nature of the gas can be ascertained from the fact that the blue - shifted emission is sharply cut out on one side by the foreground damped ly@xmath1 absorption from the galaxy .
the simple double humped structure of the emission line may arise when ionizing radiation directly impacts a filament of hi gas , `` reflecting '' part of the energy back in the form of ly@xmath1 , as opposed to the more highly modified line profiles expected for ly@xmath1 emerging through multiple scattering from the depth of an intact halo .
the ionized photons produced by the stars in that galaxy can account for the entire observed flux in ly@xmath1 if we assume that about 50% of the photons escape the halo to hit the external hi . in the absence of other possible sources of ionization this may be an actual observation of a massive escape of stellar ionizing radiation from an interacting galaxy .
the presence of an agn , as an alternative source of ionizing photons and reduced opacity , can not be ruled out , but we have no positive evidence for it . in the absence of an agn , the ionizing photons will have to escape through actual holes in the hi cocoon of the galaxy , or perhaps from a partly stripped , tidal tail surrounded by regions of lower lyman limit opacity .
the object is also remarkable in that part of the fluorescent ly@xmath1 emerges from a thin filamentary structure of in - falling gas connected to the galaxy .
the filament may be a plausible detection of a cold accretion inflow , as it does not seem to line up with any stellar features ( @xcite ) .
2-d spectrum ( lhs panel ) , hst acs b - band ( central panel ) , and v - band image ( rhs panel ) .
the image size is @xmath13 .
the spectrum shows a filamentary patch of ly@xmath1 emission . in the broad band images , several blueish galaxies are visible , together with patchy or filamentary emission features . of these
, t1 appears to be a continuum feature , whereas t2 and t3 may be dominated by ly@xmath1 emission .
t2 could be a tidal tail or a turbulent stripped wake behind a galaxy moving through the gaseous halo .
recent star formation may explain the apparently large equivalent width ( @xcite).[2ndsys ] ] the second object is at z=2.63 , and the spectrum again shows a complex and extended , partly filamentary emission line ( @xcite ; left - most panel in fig .
[ 2ndsys ] ) .
a second filament ( `` t2 '' central panel ; mostly outside of the spectrograph slit ) appears in the b - band image but not in the v - band , suggesting that its flux may also be dominated by ly@xmath1 line radiation , however , with a very large equivalent width .
the underlying continuum sources here again exhibit several diffuse emission features that may be of tidal origin .
most galaxy redshifts can not be established with absolute certainty because of the faintness of the objects , and the multiple published photometric redshifts for several objects are mutually incompatible .
we propose that this may be due to the presence of a very young ( a few @xmath14yr ) stellar population in some of the objects ( @xcite ) . in the current case
we can not discern direct evidence for escaping ionizing radiation , but the filamentary features suggest that the galaxies have interacted in ways that produced an extended stellar distribution , with possible tidal tails or stripped wakes exhibiting signs of recent star formation .
the large ly@xmath1 equivalent widths inferred and the fact that the stars appear to be forming in situ in a metal - poor gaseous halo suggests that these may be hot stars with enhanced yields of ionizing photons ( e.g. , @xcite , @xcite ) .
the location of the stars outside of the optically thickest regions of the individual galaxies would also favor the escape of ionizing radiation .
thus , this object meets several of the hypothetical criteria outlined above for the galactic sources of ionizing photons .
we have discussed some astrophysical conditions that could turn galaxies into sources of ionizing radiation during the epoch of reionization , and we described two recently discovered ly@xmath1 emitting galaxies that satisfy several of these conditions , including an apparently leaky gaseous halo , and possible extragalactic , intra - halo star - formation from a metal poor reservoir of gas ( apparently in the form of tidal tails or turbulent wakes ) .
the connection of these features to galaxy interactions suggest that , as the merger rate increases with redshift , such objects will be much more common at the epoch of reionization .
the extended @xmath0 emitters discussed here may just be the lower redshift remnants of this population .
in contrast , the symmetric , compact ly@xmath1 emitters or lyman break galaxies represent the bulk of the @xmath0 star - forming galaxies . in the absence of recent interactions , these objects , with their more spatially concentrated star - formation , more mature stellar populations , and their ionizing radiation from a dominant central source being trapped by an inviolate gaseous halo , may be less likely to contribute to the ionization of the universe . in neither one of the two peculiar ly@xmath1 emitter cases
can we exclude the presence of an agn , which could provide alternative explanations for both the high ly@xmath1 widths observed , and the escape of ionizing photons .
however , taking into account that there is already another qso in the same ( small ) survey volume , the assumption , that all three objects could reflect photonionization by internal qsos , would raise the number density for qsos at @xmath2 to @xmath15mpc@xmath9 .
such a high value in itself may have interesting consequences for the budget of ionizing photons .
i would like to thank my collaborators george becker , martin haehnelt , jean - rene gauthier , swara ravindranath , and wal sargent , and acknowledge useful discussions with sebastiano cantalupo , bob carswell , hsiao - wen chen , jeff cooke , li - zhi fang , pat mccarthy , masami ouchi , and francois schweizer .
i further thank the staff of the las campanas and the keck observatories for their help with the observations , the ioa in cambridge and the raymond and beverley sackler distinguished visitor program for hospitality and support , and the nsf for funding through grant ast-1108815 . | known populations of qsos appear to fall short of producing the ionizing flux required for re - ionizing the universe .
the alternative , galaxies as sources of ionizing photons , suffers from the problem that known types of galaxies are almost completely opaque to ionizing photons . for reionization to happen ,
either large numbers of ( largely undiscovered ) sources are required , or the known populations of galaxies need to have had a much larger escape fraction for ionizing radiation in the past .
we discuss recent discoveries of faint @xmath0 lyman @xmath1 emitters with asymmetric , extended ly@xmath1 emission regions , which apparently are related to interacting galaxies .
the unusually shaped ly@xmath1 line profiles and the underlying stellar populations of these objects suggest the presence of damaged gaseous halos , infall of gas , tidal or stripped stellar features and young populations of hot stars that would all be conducive to the release of ionizing radiation into intergalactic space . as galaxy interactions and mergers
increase with redshift , these effects can only become more important at earlier times , and so these interacting @xmath2 objects may be late , lower redshift analogues of the sources of reionization .
address = carnegie observatories , 813 santa barbara street , pasadena ca 91101 , usa |
the isospin violating strong decay @xmath3 allows access to light quark mass ratios . at lowest order of chiral perturbation theory ( chpt )
the amplitude is proportional to the light quark mass difference @xmath6 and may be written as @xmath7 where @xmath8 is the pion decay constant , @xmath9 and @xmath10 . at higher order of chpt
it has been found that final state pion interaction contribute to the decay width@xcite@xmath11@xcite .
the decay width scales as @xmath12 , where @xmath13 , @xmath14 , and the decay width @xmath15 and @xmath16 are calculated in the dashen limit@xcite .
this scaling works under the pre - requisite that @xmath15 is understood reliably . to test this , theoretical predictions and experimental measurements of pion kinematical distributions
may be compared in a dalitz plot , where the axes are defined as @xmath17 , @xmath18 . here
@xmath19 , @xmath20 and @xmath21 denote the kinetic energies of @xmath22 , @xmath23 and @xmath24 in the rest frame of the @xmath0 meson , and @xmath25 .
the standard way to parametrize the dalitz plot density is a polynomial expansion around the center point : @xmath26 where @xmath27 are the dalitz plot parameters .
the experimental results are dominated by kloe with a dalitz plot containing 1.34@xmath2810@xmath29 events @xcite .
this result shows a significant deviation of parameters @xmath30 and @xmath31 in comparison to the theoretical predictions based on chpt .
it is therefore important to perform an independent measurement , which is one of the aims of the wasa - at - cosy experiment .
in 2008 and 2009 wasa - at - cosy@xcite measured @xmath32 reaction at beam energy 1 gev , collecting @xmath33 and @xmath34 @xmath0 mesons respectively . the missing mass with respect to @xmath35
is used to tag the @xmath0 meson ( fig .
[ mm(he ) ] left ) .
in addition two tracks of opposite charge are required in the mini drift chamber in the angular range @xmath36 .
furthermore two photons with an invariant mass close to @xmath24 are required .
the @xmath37 reaction is reduced by imposing conditions on the missing mass calculated for @xmath38 and the missing mass calculated for @xmath39 .
the preliminary analysis yields 149 000 @xmath3 candidates from the 2008 data , shown in fig . [ mm(he ) ] right .
the experimental resolution is better for the @xmath0 four - momenta from @xmath35 compared to the information derived from the @xmath0 decay products .
therefore a kinematical fit for the reaction @xmath40 has been used with @xmath41 observables fixed and a cut on the 1% level of the probability density function . to estimate the @xmath0 content in each dalitz plot bin
, a four - degree polynomial fit is performed over the background region and the fitted polynomial is subtracted in the signal region .
the preliminary experimental results for the @xmath42,@xmath43 projections of the dalitz plot are compared in fig .
[ xyprojection ] to monte carlo simulations of the @xmath3 weighted with the tree - level prediction ( equation 1 ) .
the measurement of the @xmath45 reaction was conducted in 2008 and in 2010 at beam kinetic energy 1.4 gev . the collected sample of data yields about 10@xmath46 produced @xmath0 mesons .
two protons were registered using plastic scintillator detectors and straw tube trackers and the charged pions were detected with plastic scintillator detectors and the mini drift chamber .
two protons were used to tag the @xmath0 meson in the missing mass plot showed in fig .
[ mz ] left ( here we present data only from one run ) .
[ mz ] the two @xmath47 originating from the @xmath24 meson decay were registered in the electromagnetic calorimeter .
the invariant mass of these@xmath47 is required to be close to the mass of the @xmath24 ( fig .
[ mz ] middle ) . requiring two @xmath47 in coincidence with the two protons
gives the missing mass as shown in fig .
[ mz ] right .
the work for both _ pd _ and _ pp _ data will be continued in order to obtain two independent determinations of the dalitz plot density for the @xmath3 .
this includes estimating systematical errors as well as tuning monte carlo simulation .
99 c. roiesnel , t. truong , _ nucl phys b _ * 187 * ( 1981 ) 293 .
j. gasser , h. leutwyler , _ nucl phys b _ * 250 * ( 1985 ) 539 .
r. dashen , _ phys rev _ * 183 * ( 1969 ) 1245 .
kloe collab .
( f. ambrosini _
et al_. ) , _ jhep _ * 05 * ( 2008 ) 006 , arxiv:0801.2642v2 [ hep - ex ] .
wasa - at - cosy collaboration , arxiv : nucl - ex/0411038 , ( 2004 ) . | one of the objectives of the physics programme of the wasa - at - cosy facility is to study the isospin violating @xmath0 hadronic decays into @xmath1 systems driven by the term of qcd lagrangian which depends on the d and u quark mass difference .
these studies can be made in terms of the dalitz plot parameters describing the density population which is proportional to the square of the amplitude @xmath2 .
this contribution describes the current status of the analysis of the @xmath3 decay in the @xmath4 and as well in the @xmath5 reaction with wasa - at - cosy . |
mkn 501 and mkn 421 are two nearby bl - lac type objects at z = 0.034 and 0.030 , respectively .
they are well studied sources in radio , optical and x - ray bands being among the brightest bl - lac objects at all wavelengths .
mkn 421 was the first source to be detected at tev energies , followed by mkn 501 . in the radio band ,
both sources show a kpc scale morphology and a total radio power consistent with the expectation of unified scheme models that bl - lac type sources are fr i galaxies oriented at small angles to the line of sight .
we present here new space vlbi observations of mkn 501 and vlbi data of mkn 421 .
mkn 501 was observed at 18 cm on august 4th , 1997 and april 8th , 1998 in a space vlbi project using the halca satellite and 12 ground stations .
mkn 421 was observed at 6 cm in july 1995 with the vlba and at 18 cm in february 1996 with the global vlbi .
the data of both sources were correlated in socorro ( nm - usa ) and reduced with the aips package . a hubble costant of 50 km sec@xmath1 mpc@xmath1 is assumed throughout .
_ mkn 501_. at parsec resolution mkn 501 shows a one - sided jet which , at @xmath0 20 mas from the core , changes its orientation from @xmath0 140@xmath2 to @xmath0 30@xmath2 . from the analysis of the map ( fig .
1-left ) we note a clear change in the jet structure : at the beginning the jet is resolved and its brightness is centrally peaked , but at @xmath0 8 mas from the core , it becomes limb - brightened with the maximum of the surface brightness on both sides of the parsec scale jet .
we interpret this observational result as indication of a change in the jet physical properties at @xmath0 8 mas ( 7.4 pc ) from the core .
the limb - brightened jet is well visible for 20 - 25 mas ( @xmath0 20 pc ) .
this result is in agreement with the polarization properties of this source presented by aaron et al .
( present proceedings ) . at a larger distance from the core ( @xmath0 20
pc ) , the jet shows a large change in its position angle and a dramatic expansion . in our low resolution map ( fig .
1-right ) the jet is visible for more than 100 mas and appears still edge - brightened .
no evidence of a helical structure is visible in our maps .
_ mkn 421_. at 6 cm mkn 421 shows a core emission and a one - sided jet .
the jet is well collimated at the beginning , but at @xmath0 5 mas ( 4 pc ) from the core it shows many wiggles and changes in the position angle ( fig .
this complex structure is confirmed by the 18 cm map ( fig .
2-right ) at lower resolution and on a larger scale . at @xmath0 20 mas from the core
, the parsec scale jet shows a dramatic expansion , its transversal size being comparable with its longitudinal dimension .
present data do not have enough resolution or sensitivity to distinguish between a centrally peaked or a limb - brightened jet .
we used the available vlbi data to derive information on the jet velocity and orientation . assuming that parsec scale jets are intrinsically two - sided and symmetric , we can constrain the jet velocity ( @xmath3 = v / c ) and inclination to the line of sight ( @xmath4 ) from the observed jet asymmetry . for * mkn 501 * the jet / counterjet ratio is 200 , which leads to @xmath3 cos@xmath4 0.79 . moreover , from the known correlation between the core power and the unbeamed total radio power ( giovannini et al . , 1988 ) we can infer the expected intrinsic core radio power from the observed total radio power . comparing
the expected and the measured core radio power , we obtain that mkn 501 has to be oriented at @xmath4 26@xmath2 with @xmath3 in the range 1 - 0.88 .
we produced identical maps from the first and second epoch space vlbi data of mkn 501 at 18 cm .
a comparison of the two maps suggests the existence of a possible proper motion between the two epochs with apparent velocity @xmath0 6.7c .
we are aware that to measure a proper motion with only two epoch data may give unreliable results , however we note that the present data are of good quality and the structures in the two maps are in good agreement .
more space vlbi observations have been requested to confirm this result .
such an apparent velocity implies that the real jet velocity has to be 0.989c and the jet has to be oriented at an angle smaller than 17@xmath2 .
if we assume that the bulk and pattern jet velocity are comparable , we can compare the observational constraints previously discussed with the measured proper motion . from these results we derive that mkn 501 is oriented at @xmath0 10@xmath2 - 15@xmath2 with a velocity in the range 0.990 - 0.999c .
it implies high values of the lorentz factor : @xmath5 7 - 22 , but a relatively low doppler factor : @xmath6 1.3 - 5.6 . from the jet sidness and core dominance of * mkn 421
* we derive for this source an inclination angle @xmath4 @xmath7 30@xmath2 and @xmath3 in the range 1 - 0.84 . the measured proper motion for mkn 501 and
the derived intrinsic jet velocities suggest that bl - lac type objects may have parsec scale jets with intrinsic velocities of the same order of parsec scale jets in radio quasars .
thanks to the high resolution provided by the vsop we have imaged the parsec scale jet of mkn 501 in great detail .
it shows a change in its physical properties at 8 mas from the core : the innermost jet is centrally brightened , but at @xmath0 8 mas from the core it becomes limb - brightened . such a structure is expected by the _ central spine shear layer _ jet model ( laing , 1996 ) .
the observational data discussed here imply that the shear layer is not visible at the jet beginning but appears at @xmath0 7 pc from the core corresponding to a deprojected distance of 40 - 45 pc assuming @xmath4 = 10@xmath2 . at @xmath0 20 - 25 mas from the core
, the jet shows a prominent bend and a rapid expansion .
a similar expansion is found also in mkn 421 , where the jet dramatically expands at about the same distance from the core .
we interpret this result as evidence of a strong interaction with the surrounding medium which affects the jet dynamics , causing the jet velocity decrease observed in fr i radio galaxies .
current models of gamma ray emission and intra - day variability suggest for these sources that in the innermost region ( 0.003 - 0.03 pc ) the jet should be oriented at @xmath4 5.7 @xmath2 and should have a lorentz factor @xmath8 10 which implies @xmath9 0.995 and @xmath6 10 ( see spada et al .
present proceedings and salvati et al . ,
these results seem to be in contrast with the derived constraints from vlbi radio data , however we note that the regions where the radio emission ( _ the core region _ ) and the gamma and x - ray emission are produced , are not coincident , the former being at a larger distance from the active core than the latter .
we estimate that the _ radio core _ has a size 0.05 pc
. therefore radio and gamma ray data are in agreement if when travelling between the distance of 0.003 pc to 0.05 pc , the jet orientation changes from @xmath0 5@xmath2 to @xmath0 10@xmath2 .
such a change in the jet orientation is plausible given the distorted morphology found on the parsec scale of mkn 501 and mkn 421 .
we acknowledge partial financial support from italian murst and asi .
we gratefully acknowledge the vsop project , which is led by the japanese institute of space and astronautical science in cooperation with many organizations and radio telescopes around the world .
the nrao is operated by associated universities , inc . under a cooperative agreement with the national science foundation . | we present here two epochs of space vlbi observations at 18 cm of the bl - lac type object mkn 501 .
thanks to the high resolution of these new data we have found that the inner jet is centrally brightened at its beginning but becomes extended and limb brightened at @xmath0 8 mas from the core . moreover a comparison between the two epochs shows the presence of a possible proper motion with apparent velocity = 6.7c .
vlbi data at 6 and 18 cm of the bl - lac type object mkn 421 are also presented .
observational data have been used to constrain the jet velocity and orientation .
# 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in |
the cosmic dark ages ended with the formation of the first stars in 10@xmath110@xmath2 @xmath3 cosmological halos at @xmath4 2030 .
besides ionizing the universe with intense uv flux and forming the first heavy elements , pop iii stars also determined the luminosities and spectra of primeval galaxies and may have been the origin of the supermassive black holes ( smbhs ) at the centers of most massive galaxies today . unfortunately , because they lie at the edge of the observable universe , individual pop iii stars will likely remain beyond the reach of direct observation for decades to come despite their enormous luminosities @xcite and new near infrared ( nir ) observatories such as _ jwst _ @xcite and the _ thirty meter telescope _ ( tmt ) .
past numerical simulations have suggested that pop iii stars form in isolation in halos , with masses of 30300 @xmath3 @xcite , but more recent calculations point to the possibility of binaries @xcite , and perhaps small swarms of less massive 2040 @xmath3 stars @xcite .
however , in spite of their increasing sophistication numerical simulations can not yet constrain the properties of the first stars , as they do not realistically bridge the gap between the formation and fragmentation of protostellar disks and their destruction by nascent stars up to a myr later .
it may soon be possible to deduce the properties of pop iii stars from their supernova ( sn ) explosions .
primordial stars from 1040 @xmath3 are thought to die in core - collapse sne while 140260 @xmath3 stars explode as pair - instability ( pi ) sne , extremely energetic thermonuclear explosions with up to 100 times the energy of type ia or ii sne @xcite .
pop iii stars above 100 @xmath3 encounter the pair instability after central carbon burning , when thermal energy creates electron - positron pairs rather than providing pressure support against collapse @xcite .
their cores subsequently contract , triggering explosive thermonuclear burning of o and si . above 140 @xmath3
the energy released completely unbinds the star , with no compact remnant . at 260 @xmath3
the core becomes so hot that alpha particles are photodisintegrated into free nucleons , consuming as much energy per unit mass as was released by all preceding burning , and the star collapses instead of exploding .
pi sne such as sn 2007bi have now been discovered in the local universe @xcite , and the recent discovery that rotating pop iii stars can die as pi sne at masses as low as 65 @xmath3 may increase the number of such explosions at high redshifts by a factor of a few @xcite . pop iii pi sne could be ideal probes of the primeval universe because they are 100,000 times brighter than their progenitors and the early galaxies in which they reside , and might be found by _
jwst _ or in all - sky surveys by _ wfirst _ or the _ wide - field imaging surveyor for high - redshift _ ( _ wish _ ) .
however , to determine if these events can be seen from earth , fully radiation - hydrodynamical simulations of pi sne must be performed in realistic circumstellar envelopes , whose opacity and interaction with the shock crucially shape the temperatures and spectra in the source frame , and thus the flux that is eventually redshifted into the nir .
lyman absorption by neutral hydrogen prior to the era of reionization , which can absorb or scatter light from these ancient supernovae out of our line of sight , must also be taken into account for explosions at the earliest epochs .
previous simulations of pi sne have not included these effects @xcite , and estimates of the redshift at which they can be detected have either been confined to @xmath5 15 ( the era of primitive galaxies rather than the epoch of first light ) or have been attempted at @xmath4 30 with only very approximate methods @xcite .
we have performed numerical simulations of the nir light curves of pop iii pi sne . in @xmath6
we describe our numerical method and in @xmath7 we present source frame light curves and spectra for 5 pi explosions . in @xmath8 nir
light curves for these events are shown for several redshifts , and we discuss their detection limits in redshift .
ccccc 150 & 72 & 1.62 & 9.0 & 0.07 + 175 & 84.4 & 1.74 & 21.3 & 0.70 + 200 & 96.7 & 1.84 & 33 & 5.09 + 225 & 103.5 & 3.33 & 46.7 & 16.5 + 250 & 124 & 2.25 & 69.2 & 37.9
our calculations are done in four stages .
first , we model the evolution and initial explosion of 150 , 175 , 200 , 225 , and 250 @xmath3 pop iii stars in the kepler code @xcite .
second , the shock is propagated through the interior of the star , its surface , and then out into the surrounding medium with the radiation hydrodynamics code rage @xcite .
we model radiation flows with grey flux - limited diffusion and evolve matter and radiation temperatures separately because they are not always in equilibrium .
this physics enables us to better capture shock breakout ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , as discussed in greater detail in @xcite .
energy deposition in the gas due to the radioactive decay of @xmath9 , which powers the light curve at later times , is also included .
we surround the star with a power - law wind profile with a total mass of 0.1 @xmath3 , in keeping with the general consensus that massive zero - metallicity stars do not exhibit much mass loss due to their pristine atmospheres @xcite .
the wind profile is smoothly joined to the h ii region density profile of a massive pop iii star calculated with the zeus - mp code @xcite .
zeus - mp also captures the ionization state of the wind due to uv radiation from the star , and therefore its opacity to radiation from the sn .
because all five stars become red hypergiants and cease to emit ionizing radiation in the final few hundred kyr of their lives , the wind completely recombines by the time they explode while the low - density h ii region beyond it remains ionized .
third , we construct spectra from the rage blast profiles with the spectrum code @xcite using atomic opacities from the los alamos oplib database @xcite .
spectrum calculates continuum and line luminosities , absorption lines imprinted on the flux by the ejecta and envelope , and doppler shifts and time dilation due to relativistic expansion .
finally , we convolve our spectra with lyman absorption by neutral clouds @xcite , cosmological redshifting , and filter response functions to calculate nir light curves . we summarize the properties of our explosions in table 1 .
bolometric source frame luminosities are shown for all five explosions in figure [ fig : lcs ] .
the shock is not visible to an external observer before breaking through the surface of the star because electrons in the intervening layers scatter the photons .
when it reaches the surface , the shock abruptly accelerates in the steep density gradient , becoming even hotter and releasing a brilliant pulse of photons with peak luminosities above 10@xmath10 erg s@xmath11 , or 2500 times the luminosity of the milky way galaxy .
the transient is mostly x - rays and hard uv , and its duration is related to the radius of the star , since photons emitted simultaneously from its poles and equator reach an observer at times that are separated by the time it takes light to cross the star .
the temperature of the shock at breakout is @xmath12 10@xmath2 k. after the initial transient subsides , the luminosities gradually decline from 10@xmath13 erg s@xmath14 to 10@xmath15 erg s@xmath14 over 3 yr as gamma rays from the decay of @xmath9 formed during the explosion downscatter in energy and diffuse out through the ejecta .
pair - instability sn light curves are powered at early times by the conversion of kinetic energy into thermal energy by the shock , and are thus far more luminous than type ia and ii sne because they have much higher explosion energies . at later times
they are primarily powered by radioactive decay , and they remain much brighter than other sne because they have synthesized more @xmath9 : up to 50 @xmath3 compared to 0.40.8 @xmath3 and @xmath16 0.3 @xmath3 in type ia and ii sne , respectively .
they are luminous for longer times ( 3 yr instead of 36 mo for type ia and ii sne ) because radiation diffusion timescales in their more massive ejecta are longer : @xmath17 here , @xmath18 is the average opacity of the ejecta , m@xmath19 is the mass of the ejecta , and @xmath20 is the explosion energy .
all five light curves slightly rebrighten at 4 months when the photosphere ( the surface from which photons stream freely into space ) recedes into the hot @xmath9 layer deep in the ejecta .
this emission peaks at @xmath12 3000 , and its luminosity increases with the mass of @xmath9 , and hence the mass of the progenitor .
rebrightening occurs later in more massive stars because more time is required for the photosphere to sink into the @xmath9 in the frame of the shock .
figure [ fig : spectra ] shows spectra for the 225 @xmath3 explosion at four times ranging from shock breakout to 3 yr . as the fireball expands it cools , and its spectral peak advances to longer wavelengths .
the breakout transient ionizes the envelope shrouding the star , as seen at 2.9 @xmath21 10@xmath1 s ( 3.4 days ) when the spectrum has far fewer absorption lines than at later times when the shock cools and can no longer fully ionize the wind . as the envelope recombines
it imprints many more absorption lines on the spectrum , especially at shorter wavelengths where line blanketing effectively shears off the spectrum .
consequently , the common practice of fitting blackbodies to bolometric light curves to approximate shock spectra can seriously overestimate the nir flux in the observer frame , because such profiles have a significant amount of flux at short wavelengths that is actually removed by the envelope and ejecta . on the other hand ,
our bolometric luminosities are an order of magnitude higher at shock breakout than previous pi sn light curves because the shock crashes into the wind and heats it , and because radiation and matter are not forced to be in thermal equilibrium in our calculations . by 3
yr most of the photons in the spectrum are below the ionization limit of hydrogen and pass through the envelope , although resonant scattering continues to imprint absorption lines on flux below this limit .
at high redshift , lyman absorption will absorb most luminosity at wavelengths far blueward of 1216 .
however , it could be possible that the fireball is brighter a little blueward of the lyman limit rather than a little redward , even with absorption . to determine in which _ jwst _ nircam filter an explosion is brightest at a given redshift , we calculate its nir signal at wavelengths above and below the lyman limit in the observer frame . at @xmath22 30
we find that although all five sne are brighter just below 1216 in the source frame , igm absorption reduces these luminosities to magnitudes below those just redward of the limit for which there is no absorption , as we show for the 225 @xmath3 pi sn in figure [ fig : lya ] .
the explosion is brighter at 1148 in the source frame ( 3.56 @xmath23 m in the observer frame ) at early times than at 1432 ( 4.44 @xmath23 m in the observer frame ) , but after igm absorption is taken into account the event is brighter at 4.44 @xmath23 m in the observer frame . in the left panel of figure [ fig : nir ]
we show nir light curves for all five pop iii sne at 4.44 @xmath23 m in _ jwst _ at @xmath22 30 , or 100 myr after the big bang .
the detection limit for this filter is ab magnitude 32 , so all five explosions will be visible for over a thousand days .
their light curves do not cross above the detection threshold until 50 days ( or 1.5 days in the source frame at this redshift ) , so the most luminous component of the explosion , shock breakout , can not be seen from earth .
the x - rays and extreme uv in the pulse are all redshifted into the uv and absorbed by neutral hydrogen .
the nir flux in the observer frame exhibits significantly more variability than the bolometric flux because of the expansion and cooling of the fireball , and this is essential to identifying these events as explosive transients .
the flux rises much more quickly than it falls , so it is easiest to detect pop iii pi sne in their initial stages .
nonetheless , they exhibit sufficient variability over likely protogalactic survey times of 15 yr to be identified at later times as well .
this variation over timescales of @xmath121000 days will also allow them to be discriminated from primeval galaxies , in spite of their probable overlap in color - color space .
although recent calculations show that roughly one pop iii pi sn will be present in any given _
deep field @xcite , far greater numbers could be found in proposed all - sky nir surveys with _
wfirst_. with a sensitivity of ab magnitude 27 at 2.2 @xmath23 m , _ wfirst _ will see such explosions at @xmath22 1520 , as we show in the right panel of figure [ fig : nir ] .
such redshifts may be optimal for detecting pi sne because of the rise of uv background fields from the first generations of pop iii stars .
this background is thought to delay primordial star formation in halos until they have grown to larger masses , forming more massive stars at slightly lower redshifts @xcite .
star formation rates extracted from large - scale cosmological simulations of early structure formation that include uv backgrounds @xcite suggest that wfirst will detect up to 10@xmath24 pi sne per year at @xmath25 .
detections of this volume will enable the first major surveys of the pop iii initial mass function ( imf ) ( see * ? ? ? * ; * ? ? ? * for discussions on strategies for detecting superluminous type iin sne in all - sky surveys ) .
our calculations definitively establish that _ jwst _ and _ wfirst _ will be able to detect pop iii supernovae at the era of first light .
in addition to unveiling the nature of primordial stars and constraining scenarios for early cosmological reionization and chemical enrichment , detections of pop iii supernovae at slightly lower redshifts ( @xmath4 1015 ) will probe the era of primitive galaxy formation , revealing their positions on the sky when they might otherwise have eluded detection by _ jwst_.
the discovery of the first cosmic explosions will be one of the most spectacular results in extragalactic astronomy in the coming decade , and open the first window on the end of the cosmic dark ages .
the authors thank the anonymous referee , whose comments improved the quality of this paper .
djw was supported by the bruce and astrid mcwilliams center for cosmology at carnegie mellon university .
deh acknowledges support from the national science foundation career grant phy-1151836 .
ah was supported by the us department of energy under contracts de - fc02 - 01er41176 , fc02 - 09er41618 ( scidac ) , and de - fg02 - 87er40328 .
sew was supported by the national science foundation grant ast 0909129 and the nasa theory program grant nnx09ak36 g .
ms thanks marcia rieke for making available the nircam filter curves , and was partially supported by nasa jwst grant nag5 - 12458 .
work at lanl was done under the auspices of the national nuclear security administration of the u.s .
department of energy at los alamos national laboratory under contract no .
de - ac52 - 06na25396 . , a. , mazzali , p. , ofek , e. o. , nugent , p. e. , kulkarni , s. r. , kasliwal , m. m. , quimby , r. m. , filippenko , a. v. , cenko , s. b. , chornock , r. , waldman , r. , kasen , d. , sullivan , m. , beshore , e. c. , drake , a. j. , thomas , r. c. , bloom , j. s. , poznanski , d. , miller , a. a. , foley , r. j. , silverman , j. m. , arcavi , i. , ellis , r. s. , & deng , j. 2009 , , 462 , 624 , j. p. , mather , j. c. , clampin , m. , doyon , r. , greenhouse , m. a. , hammel , h. b. , hutchings , j. b. , jakobsen , p. , lilly , s. j. , long , k. s. , lunine , j. i. , mccaughrean , m. j. , mountain , m. , nella , j. , rieke , g. h. , rieke , m. j. , rix , h .- w . , smith , e. p. , sonneborn , g. , stiavelli , m. , stockman , h. s. , windhorst , r. a. , & wright , g. s. 2006 , , 123 , 485 , n. h. , abdallah , jr . ,
j. , clark , r. e. h. , cohen , j. s. , collins , l. a. , csanak , g. , fontes , c. j. , gauger , a. , keady , j. j. , kilcrease , d. p. , & merts , a. l. 1995 , in astronomical society of the pacific conference series , vol .
78 , astrophysical applications of powerful new databases , ed .
s. j. adelman & w. l. wiese , 51 , j. , stiavelli , m. , oesch , p. , trenti , m. , bergeron , e. , bradley , l. , carollo , m. , dahlen , t. , ferguson , h. c. , giavalisco , m. , koekemoer , a. , lilly , s. , lucas , r. a. , mobasher , b. , panagia , n. , & pavlovsky , c. 2011 , , 738 , 123 | the first stars ended the cosmic dark ages and created the first heavy elements necessary for the formation of planets and life .
the properties of these stars remain uncertain , and it may be decades before individual pop iii stars are directly observed .
their masses , however , can be inferred from their supernova explosions , which may soon be found in both deep - field surveys by the _ james webb space telescope _ ( _ jwst _ ) and in all - sky surveys by the _ wide field infrared survey telescope _ ( _ wfirst _ ) .
we have performed radiation hydrodynamical simulations of the near infrared signals of pop iii pair - instability supernovae in realistic circumstellar environments with lyman absorption by the neutral intergalactic medium .
we find that _ jwst _ and _ wfirst _ will detect these explosions out to @xmath0 and 20 , respectively , unveiling the first generation of stars in the universe . |
in recent years there has been an extensive experimental and theoretical activity in the field of magnetoelectrics . the magnetoelectric effect ( me )
was originally predicted by curie in 1894,@xcite and describes the induction of electric polarization @xmath2 by a magnetic field and , vice - versa , the induction of magnetization @xmath12 by an electric field @xmath13 .
a phenomenological theory was developed by landau,@xcite who considered the invariants in the expansion of the free energy up to linear terms in the magnetic field . in this approach
, symmetry considerations alone can fix the form of the coupling between @xmath2 and @xmath12.@xcite based on such symmetry arguments , dzyaloshinskii proposed the antiferromagnetic ( afm ) compound cr@xmath0o@xmath1 as a first candidate for the observation of the me effect.@xcite indeed , the electrically induced magnetization ( me@xmath14 ) in cr@xmath0o@xmath1 was observed experimentally for the first time by astrov,@xcite and soon after , rado and folen@xcite measured the the converse effect , i.e the magnetic - field induced electric polarization ( me@xmath7 ) . despite the scarcity of compounds with cross - coupled magnetic and electric properties , which is related to the antagonistic requirements for the existence of
both orders,@xcite the research in the field has grown enormously .
since the discovery of me effects in cr@xmath0o@xmath1 , more than 100 magnetoelectric compounds have been discovered or synthesized.@xcite the motivation behind this activity stems from the novel and fundamental physical phenomena involved , but also for exciting potential applications , including me sensors and the electric control of magnetic memories without electric currents ( and thus joule heating).@xcite at the center of interest have been the so - called multiferroics , which are materials possessing at least two switchable order parameters , such as electric polarization @xmath2 , magnetization @xmath12 , and strain.@xcite multiferroics show spontaneous me effects in addition to those induced by external fields . despite the coexistence of ferroelectricity and magnetism , very few multiferroic materials exhibit strong coupling between @xmath2 and @xmath12,@xcite and in most cases the coupling is rather weak .
typical examples are the perovskites bimno@xmath1 and bifeo@xmath1 , where the ferroelectric transition temperature is much higher than the magnetic one.@xcite a breakthrough in the field came with the discovery of the multiferroics tbmno@xmath1@xcite and tbmn@xmath0o@xmath15,@xcite where the ferroelectricity is directly driven by the spin order .
the magnetic state in these compounds is either a spiral ( as in tbmno@xmath1,@xcite ni@xmath1v@xmath0o@xmath16,@xcite and mnwo@xmath17@xcite ) , or a collinear configuration ( as in ca@xmath1(comn)@xmath0o@xmath18@xcite ) . for the understanding of the me effects in these systems
, there has been a number of proposed microscopic mechanisms ( such as the spin - current model,@xcite the exchange - striction model,@xcite and the electric - current cancelation model@xcite ) , in addition to landau - ginzburg phenomenological theories.@xcite here we focus on the cubic magnetoelectric insulator cu@xmath0oseo@xmath1.@xcite this system crystallizes in the non - centrosymmetric space group _ _ p__2@xmath193,@xcite which allows for piezoelectricity and piezomagnetism but not for a spontaneous polarization .
the structure is a three - dimensional ( 3d ) array of distorted corner - sharing tetrahedra of copper spin @xmath20 ions .
the unit cell consists of 16 copper ions which belong to two crystallographically inequivalent groups denoted here by cu@xmath21 and cu@xmath22 ( wyckoff positions 4@xmath23 and 12@xmath24 , respectively ) .
each tetrahedron comprises three cu@xmath0 and one cu@xmath19 ion .
polycrystalline samples of this compound were studied by bos _
et al._,@xcite who reported a transition to a ferrimagnetic state at t@xmath25 k. based on neutron powder diffraction data , bos _ et al._@xcite proposed that in this state all cu@xmath0 moments prefer to align parallel to each other and anti - parallel to the cu@xmath19 moments
. when all cu spins have their full moment , this 3up-1down state corresponds to a 1/2 magnetization plateau , which is realized for h@xmath261 koe.@xcite the 3up-1down nature of this plateau was later confirmed by high field ( 14 t ) @xmath27se nuclear magnetic resonance ( nmr).@xcite the presence of a finite me coupling in cu@xmath0oseo@xmath1 is revealed by an anomaly in the dielectric constant,@xcite at the onset of the magnetically ordered state . at the same time , high resolution synchrotron x - ray powder diffraction data@xcite show that the system remains metrically cubic down to 10 k , with no anomalous change in the lattice constant through the magnetic transition .
this is also supported by infrared@xcite and raman@xcite studies in single crystals .
furthermore , the nmr study by belesi _
et al._@xcite demonstrates that there is not any measurable structural distortion even in high magnetic fields.@xcite these findings strongly suggest that the dielectric constant anomaly is not related to any magnetostructural coupling or polar structural distortions,@xcite but is rather driven by the ( primary ) magnetic order parameter .
the magnetism of cu@xmath0oseo@xmath1 below the 1/2 magnetization plateau is very special . on the theoretical side
, the reason can be understood from its non - centrosymmetric crystal structure that belongs to laue class t ( 23 ) . in cubic magnets from this class ,
the magnetic ( free ) energy contains lifshitz - type invariants , which impair the homogeneity of the magnetic ordering and cause a continuous , helical twisting of the magnetic order parameter with a fixed sense of rotation , as first predicted by dzyaloshinskii.@xcite microscopically , the origin of these couplings relies on the weak ( relativistic ) dzyaloshinskii - moriya ( dm ) antisymmetric exchange that leads to a slight twisting of neighboring magnetic moments.@xcite the best known cubic chiral helimagnets are the intermetallic compounds and alloys with the `` b20 '' structure , which belong also to the laue class t.@xcite in magnetic systems with this crystallographic structure , like mnsi,@xcite ( fe , co)si,@xcite , or fege,@xcite long - period transverse flat ferromagnetic helices are observed to form the magnetic ground state .
currently , there is a surge of interest in these systems because complex topological spin textures composed of topologicical solitons called `` chiral skyrmions '' have been predicted to exist in these systems under certain additional conditions,@xcite and have been observed in thin films.@xcite furthermore , near the magnetic ordering transitions , a complex sequence of unusual magnetic textures and properties have been observed,@xcite which are usually referred to as the `` a - phase '' . very recently , it has been found that the magnetically ordered ground - state of cu@xmath0oseo@xmath1 is in fact a long - period helix with a pitch of about 50 nm.@xcite the magnetism has also been described as very similar to that of the b20 chiral ferromagnets , and the observation of field - driven skyrmion textures stabilized in thin film single crystals has been reported by lorentz - microscopy.@xcite theoretically , the chiral long - period modulation is the expected behavior of a magnetic system with a simple vector order parameter that belongs to one of the crystallographic classes that allow for the existence of lifshitz - invariants .
the basic spin - structure is dictated by the isotropic exchange , and can be well described by a collinear ferrimagnetic order with very large exchange fields of the order of 50 - 100 t.@xcite in fact , detailed microscopic calculations based on density - functional theory@xcite show that no magnetic frustration should be present in cu@xmath0oseo@xmath1 , although the magnetic multi - sublattice structure appears to allow for geometric frustration .
the weak relativistic dm interactions are a secondary effect that causes a long - period twisting of this primary ferrimagnetic order .
this is in agreement with the symmetry analysis by gnezdilov _
et al._,@xcite who argued that dm interactions must be important to understand the magnetism in cu@xmath0oseo@xmath1 .
therefore , the magnetoelectric insulator cu@xmath0oseo@xmath1 should be understood as a chiral ferrimagnetic helimagnet , and its magnetization process in a field is that of the field - driven evolution of a chiral helix . here
we report an extensive magnetoelectric study on single crystals of cu@xmath0oseo@xmath1 .
our data demonstrate the me@xmath7 effect , as well as a finite magnetocapacitance ( mc ) which sets in at the onset of the magnetically ordered state ( @xmath3 k ) .
we probe these effects in considerable detail as a function of temperature ( t ) and applied field @xmath4 .
we also study the variation of @xmath2 and mc with respect to the relative orientations of @xmath4 with the crystallographic axes , as well as the relative orientation between magnetic and ac - electric @xmath13 fields ( for the mc measurements ) .
our me data manifest a number of anomalies which signal magnetic phase transitions . from these anomalies
we map out the magnetic phase diagram of the system in the h@xmath28-t plane , and demonstrate that there are at least two additional magnetic phases below the 1/2-plateau , the low - field and the intermediate - field phase .
we show that in the intermediate phase the field dependence of the me@xmath7 data is fully consistent with the evolution of a chiral conical helix , whose propagation vector is along @xmath4 . in the low - field phase , the vanishing of the me response as h @xmath29 0 suggests a multi - domain structure of flat transverse helices , where the propagation vectors of the different helices are pinned along some preferred axes of the system .
the transition between the low - field and the intermediate - field phase can then be ascribed to the complete alignment of these propagation vectors along @xmath4 .
the above picture of two additional phases at low fields seems to be consistent up to t@xmath30 30 k. at higher temperatures , further anomalies are observed which suggest in particular a complex thermal reorientation of the anisotropy .
we also find a very characteristic double - peak anomaly in the me@xmath7 effect in a very narrow ( 2k ) window close to @xmath31 , which might well be related to the possible presence of the `` a - phase '' mentioned above , which is typically expected close to @xmath31.@xcite on the theory side , we present a landau - ginzburg phenomenological theory which captures both qualitatively and quantitative features of the me@xmath7 effect , in the intermediate and the 1/2-plateau phase .
in particular , this theory explains the angular dependence of the me@xmath7 effect , as well as the robust sign change of the polarization p as we go from the intermediate to the 1/2-plateau phase .
in addition , it provides strong evidence that the me effect in this compound can be attributed to an exchange striction mechanism which involves the symmetric portion of the anisotropic exchange .
contrary to expectations , the influence of the dm coupling in the me effect is heavily diminished by the long - wavelength nature of the helimagnetism in cu@xmath0oseo@xmath1 .
the theory also predicts that apart from the uniform component of the electric polarization ( which is what we measure in the present experiments ) , there is also a spatially oscillating component ( on a mesoscopic scale ) that is naturally driven by the long - period helimagnetism in cu@xmath0oseo@xmath1 .
finally , a similar landau - ginzburg theory is also given for the mc measurements .
our article is organized as follows . in sec .
[ expdetails ] we provide details on our experiments which were performed at epf - lausanne . in sec .
[ magnetization ] we present our magnetization measurements .
the quasi - static me and mc measurements are presented in secs .
[ sec : meh ] and [ sec : mc ] , respectively . the theoretical interpretation of the me@xmath7 and mc data are given in secs .
[ sec : theoryme ] and [ sec : theorymc ] respectively .
finally , a brief summary and discussion of our study is given in sec .
[ sec : disc ] .
high quality single crystals of cu@xmath0oseo@xmath1 were grown by the standard chemical vapor phase method .
more details about the crystal growth can be found in ref . [ ] . for our measurements
we have used two single crystals of cu@xmath0oseo@xmath1 , which we denote in the following by `` crystal a '' and `` crystal b '' .
both crystals have a thin rectangular plate shape , with dimensions [email protected]@xmath320.4 mm@xmath33 ( crystal a ) and [email protected]@xmath320.4 mm@xmath33 ( crystal b ) .
the orientation of the crystal axes with respect to the crystal cuts is determined by single - crystal x - ray diffraction . in particular ,
the normal to the widest face of crystal a is [ -554 ] , which is about @xmath34 off from the body diagonal [ -111 ] axis , while the normal to the widest face of crystal b is the [ 100 ] axis .
oseo@xmath1 at 15 k. ( a ) the applied magnetic field ( right hand axis ) is linearly varied with time along the [ @xmath35554 ] direction , while the electric charge induced on the ( @xmath35554 ) surface is measured ( left hand axis ) .
( b ) the induced charge from ( a ) when plotted against the applied field.,scaledwidth=48.0% ] the magnetization measurements were performed in a quantum design magnetic property measurement system ( mpms ) . for the dielectric measurements electrical contacts
were painted using silver paste .
the capacitance of the crystals was measured using a hp 4284a lcr meter .
for the temperature and magnetic field dependence of the capacitance we have employed a mpms and a janis closed cycle refrigerator , equipped with an electromagnet and a magnetic field regulator .
the latter was also employed for the quasi - static magnetoelectric ( me@xmath7 ) measurements . for the quasistatic me@xmath7 measurements , a slowly varying magnetic field
is applied to the crystal at a fixed temperature and the induced electric charge q is recorded with an electrometer .
this method is described e.g. by rivera in refs . [ ] . in a typical experiment , we begin by measuring the charge q at zero field for a few seconds .
then , we ramp up the field from 0 to 7.8 koe at a constant rate ( 0.08 t / min to 0.4 t / min ) while recording the induced charge . the field
is then kept at 7.8 koe for a few seconds and then the magnetic field is decreased to zero , again linearly with time .
the same procedure can be continued for negative values of the field .
an example of this field cycling is presented in fig .
[ medemonstration](a ) , which shows an experiment at 15 k with the magnetic field along the [ -554 ] direction and the induced charge q measured on the ( -554 ) surface of crystal a. figure [ medemonstration](b ) shows the induced charge q as a function of the applied field .
a peak , accompanied by hysteresis , occurs at 700 oe .
the position of this peak corresponds to the field - induced transition observed in the magnetization process , and will be discussed in detail in sec .
[ sec : tdependenceme ] below .
finally , the quasistatic me@xmath7 measurements have been performed in the temperature range 10 - 60 k. is applied along the [ -554 ] and [ 100 ] crystallographic directions , which are perpendicular to the widest rectangular crystal faces .
( b ) the first derivatives of the magnetization loops.,scaledwidth=46.0% ]
the magnetic properties of cu@xmath0oseo@xmath1 were first studied by bos _ et al . _
@xcite in polycrystalline samples .
since then a number of magnetization data in single crystals of cu@xmath0oseo@xmath1 were reported in the literature.@xcite along the [ -554 ] ( a ) and [ 100 ] ( b ) crystallographic directions .
panels ( c ) and ( d ) show the derivative dm / dh for three representative curves in ( a ) and ( b ) respectively .
here we have corrected for the demagnetizing field h@xmath36 , i.e. h = h@xmath37h@xmath38 is the internal field.,scaledwidth=48.0% ] here we would like to map out the h@xmath28-t phase diagram of cu@xmath0oseo@xmath1 , so that we contrast it to the one obtained from dielectric measurements below ( sec . [
sec : mephasediagram ] ) . to this end
, we have taken detailed magnetization measurements as a function of t and h@xmath28 .
in addition , we have data for two different crystallographic orientations , by applying the field @xmath4 perpendicular to the widest crystal faces of the two samples , namely to the [ -554 ] direction for crystal a and to the [ 100 ] direction for crystal b. a representative set of data for these two crystallographic orientations is shown in fig . [ mhloop5k](a ) , at 5 k. at low fields the magnetization varies linearly with h@xmath28 , while a kink is observed at h@xmath39 oe , which is most clearly seen by a peak in the first derivative of the magnetization ( see fig . [ mhloop5k](b ) ) .
this peak indicates a field - induced transition , which was originally reported in polycrystalline samples,@xcite and was also observed in single crystals as well.@xcite at higher values of the magnetic field ( [email protected] - 2 koe at 5 k ) , the magnetization reaches a value of 0.53 @xmath41 per cu for both crystallographic directions .
this moment is consistent with the 3up-1down ferrimagnetic state that was originally proposed by bos _
et al._,@xcite and recently confirmed by high - field nmr measurements.@xcite in agreement with previous results,@xcite the magnetization reaches this 1/2-plateau earlier when the field is applied along a cube edge , as compared to a space diagonal direction , i.e. , h@xmath42}_{\text{plateau}}<$]h@xmath43}_{\text{plateau}}$ ] .
oseo@xmath1 , as follows from the anomalies in the magnetization measurements shown in fig .
[ mhderivdemcor].,scaledwidth=44.0% ] the evolution of the magnetization process with temperature is presented in figs .
[ mhderivdemcor](a ) and ( b ) , for the [ -554 ] and [ 100 ] crystallographic orientations , respectively . to highlight some of the features we also show in ( c ) and ( d ) the differential susceptibilities dm / dh . here
, all data are plotted against the internal magnetic field , h = h@xmath37h@xmath38 , where the demagnetizing field h@xmath38 has been calculated following ref .
[ ] . to map out the h - t phase diagram of cu@xmath0oseo@xmath1 we follow the t - dependence of the two main characteristic anomalies in the magnetization process , namely , the position h@xmath44 of the field - induced transition and the onset of the 1/2-plateau h@xmath45 .
collecting h@xmath44 and h@xmath45 for both crystallographic directions , we obtain the h - t phase diagram shown in fig .
[ htdemcorrn2 ] .
we first remark that there are two temperature regimes with qualitatively different features , namely above and below 30 - 35 k. in the low - t region we find that h@xmath46 } < $ ] h@xmath47}$ ] , while h@xmath42}_{\text{plateau}}<\text{h}^{[-554]}_{\text{plateau}}$ ] , similarly to the 5 k data discussed above .
these findings suggest that the magnetization prefers the body diagonals in the low - field phase , but the [ 100 ] axes above the file - induced transition . in the high - t region
, we can first observe that @xmath48}_{\text{plateau}}\simeq\text{h}^{[-554]}_{\text{plateau}}$ ] and that they both show an anomalous t - dependence .
the second observation is that , in addition to h@xmath44 and h@xmath45 , there appears a third characteristic anomaly , denoted by h@xmath49}$ ] .
this corresponds to a second peak in dm / dh which is visible for t@xmath5027 k ( fig . [ mhderivdemcor](d ) ) .
it is noteworthy that in the same t - range , the [ 100 ] data show a more pronounced hysteresis , which is absent from the [ -554 ] data .
we also note that h@xmath49}$ ] follows quite closely the t - dependence of h@xmath51}$ ] ( see fig .
[ htdemcorrn2 ] ) .
altogether , the data clearly reveal different anisotropic properties compared to the low - t region , and suggest either a coexistence region ( `` phase separation '' ) between h@xmath44 and h@xmath52 , or a separate phase whose nature needs to be investigated further . the magnetic properties in the low - t region are readily interpreted by the transformations of a ground - state composed of different domains of flat ferrimagnetic helices .
these helices have their propagation directions along the easy axes of the cubic system , i.e. the magnetic moments are rotating in the plane normal to such an axis . since the single - ion magnetic anisotropy is absent for in the present s=1/2 compound , the leading anisotropy in this cubic system should be of the exchange type . depending on the sign of this anisotropic exchange ,
the easy axes are along [ 111 ] or [ 100].@xcite the ground - state may form a thermodynamical multidomain state , composed of these helices , owing to magnetoelastic effects , similar to stable domain states in ordinary antiferromagnets .
first - order magnetization processes transform this multidomain state into a single - domain of a conical helix that propagates in the direction of the applied field and displays a net magnetization.@xcite the details of this process depend on the orientation of the field with respect to the easy axes and can be hysteretic . as will be shown below by considering the possible me coupling terms in the phenomenological theory , for a cubic magnetoeletric helimagnet like cu@xmath0oseo@xmath1 ,
the conical helices display a net dielectric polarization , while the effect from the flat helices averages out .
therefore , the magnetically driven reorientation process may proceed via an intermediate domain state that is composed of flat helices without a net dielectric polarization and the polarized field - driven conical helix . in that case
, it is the long - range electrostatic effect that stabilizes a multidomain state of magnetic helices .
the magnetization curves suggest that the transformation process into the single - conical helix is concluded close to the peak of the @xmath53 curves , see fig .
[ mhderivdemcor ] ( b ) . for higher fields
, the single - domain conical helix evolves toward the collinear ferrimagnetic state , which is achieved by a continuous process at @xmath54 .
the t - dependence of @xmath54 depicted in fig .
[ htdemcorrn2 ] , which should trace the evolution of the magnetic anisotropy with temperature , is clearly anomalous . at low temperatures
the effective easy axes in the cubic system should be of @xmath55 $ ] type at high fields .
as there exists a net dielectric polarization in the conical and collinear phase , this anisotropy may be affected by the magnetoelectric couplings and may not reflect the purely magnetic anisotropic couplings which control the behavior of the flat helices at low magnetic fields .
our survey of the magnetic phase diagram suggests that the magnetic structure is simple for fields @xmath56 as being that of a conical helix that is transformed into a collinear state by increasing magnetic field at low temperatures , where also a substantially anisotropic magnetic behavior is observed .
however , the magnetic spin - structure and transformation processes at higher temperatures and close to the magnetic ordering transition will require further experimental and theoretical studies .
the t - dependence of the dielectric constant @xmath57 of cu@xmath0oseo@xmath1 has been reported by bos _
et al._@xcite from capacitance measurements on polycrystalline samples , and by miller _
et al._@xcite following an analysis of far - infrared measurements on single crystals .
figure [ er111 ] shows our own data from capacitance measurements in crystal a with a parallel plate geometry .
the experiment was performed at zero magnetic field with an ac - electric field along the [ -554 ] direction .
the frequency was 9 khz and the voltage 1 v. we have repeated the measurement at various frequencies in the range 20 hz-1 mhz but no significant frequency dependence was observed .
the dielectric loss factor ( @xmath58 ) was of the order of @xmath59 in our measurements .
the t - dependence of @xmath57 was also measured for crystal b but the results are identical .
our single - crystal data are in good qualitative agreement with the results reported in refs .
[ ] and [ ] . are smaller by a factor of two compared to the values reported in polycrystalline samples,@xcite and are closer to the ones reported in ref . .
second , the drop of @xmath60 observed for t@xmath6130 k as well as in the high - t range , is slower here as compared to the data presented in ref . .
] the most important feature is the enhancement of the dielectric constant below @xmath3060 k ( see fig . [ er111 ] ) , which marks the onset of an additional contribution which superimposes the high - t curve
. since this enhancement coincides with the onset of magnetic ordering , it signals the presence of a magnetoelectric coupling mechanism in cu@xmath0oseo@xmath1 .
here we present direct evidence for the me@xmath7 effect in cu@xmath0oseo@xmath1 .
namely , that an electric charge can be induced at the surfaces of the crystals by an applied magnetic field ( without an applied electric field ) . we shall also demonstrate that this phenomenon sets in at the onset of magnetic ordering . in the following
, the measured polarization actually corresponds to @xmath62 , where s is the surface on which we measure the induced charge @xmath63 , and the unit vector @xmath64 is vertical to that surface .
direction at various temperatures .
the magnetic field is also applied along the [ -554 ] orientation .
the inset shows a magnified view of the induced polarization at 56 and 57 k. ( b ) the t - dependence of the induced polarization at the position of the field - induced transition p@xmath65 ( open circles ) , and at the plateau -p@xmath66 ( filled circles ) taken from ( a ) . also included the t - dependence of the second power of the magnetization m@xmath67 at the position of the plateau for crystal a ( half - filled circles ) and crystal b ( half - filled squares).,scaledwidth=40.0% ] the me@xmath7 effect in cu@xmath0oseo@xmath1 was measured in the range 15 - 60 k following the quasi - static method described in section [ expdetails ] .
figure [ newp111vsthpar111](a ) shows the magnetic field induced electric polarization in the [ -554 ] direction for various temperatures .
the magnetic field h@xmath28 was applied along the [ -554 ] direction which is perpendicular to the widest crystal face ( see inset of fig .
[ newp111vsthpar111](b ) ) .
the data at the lowest temperature ( 15 k ) show a net polarization only for h@xmath69 350 oe ( see also fig .
[ medemonstration ] ) .
this finding is consistent with the low - field picture of multi - domain flat helices where the net polarization averages out .
we also observe that the polarization reaches a peak at about 700 oe , which suggests that the field - induced reorientation of the domains is concluded at this point .
this is also consistent with the fact that the position of this peak ( point `` a '' in the following ) is very close to the field - induced transition in the magnetization measurements ( fig .
[ mhloop5k ] ) . at higher fields , p@xmath70 drops quickly with h@xmath28 and
saturates to a negative value as soon as the system enters the @xmath71-plateau ( fig . [ mhloop5k ] ) .
this change of sign between `` point a '' and the onset of the 1/2-plateau is a very robust feature in our measurements , and , as we show in sec .
[ sec : theoryme ] , it is in fact a characteristic fingerprint of the evolution of a conical helix under a magnetic field .
direction at various temperatures . here
the magnetic field is perpendicular to [ -554 ] .
( b ) the t - dependence of the polarization data shown in ( a ) at the field - induced transition p@xmath65 ( open squares ) , and at the plateau p@xmath66 ( filled squares ) .
we also include the corresponding data ( rescaled by a factor of @xmath71 ) from fig .
[ newp111vsthpar111](b ) where @xmath4 was parallel to [ -554 ] ( circles).,scaledwidth=40.0% ] at higher t s , the overall field dependence remains the same , but the induced polarization is gradually decreasing in magnitude .
a finite polarization can be measured up to 59 k ( the anomaly at the field - induced transition can be observed up to 57 k ) which corresponds to the magnetic ordering transition .
this immediately tells us that the magnetoelectric effect is driven by the ( primary ) magnetic order parameter .
interestingly , a second anomaly is observed in the p@xmath70 vs. h curves as we approach t@xmath44 .
this anomaly ( point a@xmath72 in the inset of fig . [ newp111vsthpar111 ] ) could be observed only between 56 - 57 k. despite the fact that the induced charge is rather small in this t - range , this feature was reproducible between different runs .
the existence of this second anomaly may signal the onset of a third phase in a very narrow t - window close to t@xmath44 , and may well be related to the presence of the so - called `` a - phase''.@xcite the t - dependence of the induced polarization at the field - induced transition , p@xmath65 , and at the plateau , p@xmath66 , are presented in fig .
[ newp111vsthpar111](b ) .
apart from the sign difference , we can also observe that p@xmath73 shows a much stronger t - dependence than p@xmath65 .
furthermore , we should point out here that the t - dependence of p@xmath66 scales quite well with the second power of the magnetization m@xmath67 measured at the position of the plateau ( see fig .
[ newp111vsthpar111](b ) ) .
this happens because the leading order free - energy term responsible for the me@xmath7 effect scales quadratic with the magnetic order parameters ( sec .
[ sec : theoryme ] ) .
we have also studied the t - dependence of p@xmath70 when @xmath75[-554 ] , i.e. , on the crystal plate ( see inset of fig . [ newp111vsthperp111](a ) ) , and the results are presented in fig .
[ newp111vsthperp111 ] .
an important difference compared to the data shown in fig .
[ newp111vsthpar111 ] , is that here the induced polarization has changed its sign and the magnitude is halved .
this feature will be explained below in sec .
[ sec : theoryme ] .
we should remark here that the field - induced transition and the plateau are observed at lower magnetic fields in the measurements of fig .
[ newp111vsthperp111](a ) as compared to the data presented in fig .
[ newp111vsthpar111 ] ( a ) .
this is simply due to the lower value of the demagnetizing field for the in - plane configuration , as opposed to the out - of - plane orientation shown in fig .
[ newp111vsthpar111 ] . beside these differences , both p@xmath65 and p@xmath66 show the same t - dependence with that in fig .
[ newp111vsthpar111](b ) .
this is illustrated in fig .
[ newp111vsthperp111](b ) where we have included the data from fig .
[ newp111vsthpar111](b ) ( rescaled by a factor of -2 ) .
we have studied the angular dependence of the me@xmath7 effect for crystals `` a '' and `` b '' at 15 k. the me@xmath7 effect is strongly anisotropic and the direction of the induced polarization can be controlled by rotating the applied magnetic field . ,
measured at 15 k and for various directions of the magnetic field with respect to the [ -554 ] axis . at @xmath76
the magnetic field is parallel to [ -554 ] , while for @xmath77 the magnetic field lies on the ( -554 ) plane .
( b ) the angular dependence of the induced polarization at the plateau , at `` point a '' , and at `` point b '' .
the solid lines are fits to eq .
( [ eqn : akatak]).,width=325 ] the angular dependence of the me@xmath7 effect for the ( -554 ) crystal plate is presented in fig .
[ pvstheta111katak ] . here
we measure the polarization induced along [ -554 ] while we rotate the magnetic field from out - of - plane to in - plane , as shown in fig .
[ pvstheta111katak ] .
the induced polarization shows an anomaly at the field - induced transition ( `` point a '' ) at low fields .
the corresponding amplitude is maximum when the magnetic field is along [ -554 ] .
the angular dependence of the induced polarization at `` point a '' and at the 1/2-plateau can be both captured by our phenomenological theory of sec .
[ sec : theoryme ] ( solid lines in fig . [
pvstheta111katak](b ) ) . apart from the anomaly at
point a " , a second anomaly ( marked as point b " in fig . [ pvstheta111katak ] ) is observed between 30@xmath78 and 60@xmath78 and between 110@xmath78 to at least 125@xmath78 . in the region between 60@xmath78 and 110@xmath78 this feature
can not be observed since the field dependence of the induced polarization flattens out and merges with the saturated behavior at the @xmath71-plateau .
the appearance of two anomalies ( a and b ) must be related to a more complex domain reorientation process in this parameter region , which requires further investigations . , measured at 15 k and for @xmath4 rotating on the ( 100 ) plane ( with @xmath79 $ ] at @xmath80 ) .
( b)-(c ) .
the angular dependence of the induced polarization at the plateau and at point a " .
the solid lines are fits to eq .
( [ eqn : p100a]).,width=325 ] the angular dependence of the me@xmath7 effect for `` crystal b '' was studied in two different configurations . in the first configuration ( see fig . [ mevsphi100horiz](a ) ) we measure the field - induced polarization along [ 100 ] while we rotate the magnetic field in the ( 100 ) crystal plane . for @xmath81=0@xmath78
the magnetic field is along the [ 010 ] axis .
we see that we can switch the direction of the induced electric polarization by rotating the magnetic field in the ( 100 ) plane .
in particular , p@xmath82 almost vanishes when the magnetic field is parallel to the cubic edges [ 010 ] , [ 001 ] , and is maximum when it is along the cube face diagonals .
the field dependence of p@xmath82 is similar to the results obtained from the ( -554 ) platelet ( figs .
[ newp111vsthpar111 ] , [ newp111vsthperp111 ] , and [ pvstheta111katak ] ) .
some finite polarization is observed below the field - induced transition ( `` point a '' ) , while at higher fields the slope of the p _ vs. _ h curves changes sign and finally p@xmath82 saturates as soon as the system enters the 1/2-plateau .
the angular dependence of p@xmath82 at `` point a '' and at the plateau are displayed in fig .
[ mevsphi100horiz](b ) and ( c ) respectively .
both curves follow a sinusoidal dependence of the form @xmath83 , which can be captured by our phenomenological theory ( sec . [
sec : theoryme ] ) . , measured at 15 k and for @xmath4 rotating from [
100 ] ( @xmath76 ) to [ 03 - 7 ] ( @xmath77 ) .
( b ) the angular dependence of the induced polarization at the plateau , and ( c ) at point a " .
the solid lines are fits to eq .
[ eqn : p100b].,width=325 ] in the second configuration , we measure the induced polarization along [ 100 ] while we rotate the magnetic field from out - of - plane to in - plane , as shown in fig .
[ pvstheta100katak](a ) .
the polarization p@xmath82 is almost zero when the field is along [ 100 ] , and rises with the angle @xmath84 between [ 100 ] and the magnetic field .
figure [ pvstheta100katak ] shows the angular dependence of p@xmath82 at the plateau ( b ) and at `` point a '' ( c ) , and they both scale as @xmath85 .
this dependence will also be explained in sec .
[ sec : theoryme ] .
the field dependence of the magnetocapacitance ( mc ) was first studied in polycrystalline samples of cu@xmath0oseo@xmath1 by bos _
et al_. @xcite here we present our single crystal data for magnetic fields between 0 and 4 t and in the t - range 4.2 - 60 k. as we are going to show , the magnetocapacitance [ mc ] ( _ a)= exhibits a rather strong t - dependence and quite anisotropic properties . in particular , the magnitude of mc depends on the orientation of the magnetic field with respect to the crystallographic axes but also on the relative orientation between the magnetic @xmath4 and the ac - electric field @xmath86 .
oseo@xmath1 at 4.2 k for various orientations of the applied magnetic field h@xmath87 in respect to the crystallographic axis . the inset shows the mc data for crystal a plotted together with the derivative of the magnetization and the induced polarization p@xmath70 , all measured with the magnetic field applied parallel to [ -554].,scaledwidth=42.0% ] figure [ mc4p2k ] shows some representative mc data at 4.2 k for three different directions of @xmath4 with respect to the crystallographic axes , and with @xmath89 .
significant mc is observed which is dominated by a peak occurring slightly before the system enters the @xmath71-plateau ( see inset of fig .
[ mc4p2k ] ) .
the amplitude of the mc at the position of the peak is maximum when @xmath4 is along [ -554 ] , and is practically zero when it is directed along [ 100 ] . at higher magnetic fields where the system is already in the 1/2-plateau , the mc saturates to small negative values in all directions . in the inset of fig .
[ mc4p2k ] , apart from the mc data for crystal a we have also included the polarization data p@xmath70 , as well as dm / dh ( from fig .
[ mhloop5k](b ) ) for the same crystal , to demonstrate that the anomalies found in the magnetization measurements track the ones observed in the dielectric measurements , which highlights the strong me coupling in cu@xmath0oseo@xmath1 . .
the inset shows a magnified view of the mc at 4.2 k at low fields , where a small shoulder ( `` point a '' ) accompanied with some hysteresis is observed .
( b ) the magnitude of the mc at three characteristic values of the magnetic field , i.e. , at points `` a '' and `` b '' , and at the onset field of the 1/2-plateau ( rescaled by a factor of -6.6).,scaledwidth=40.0% ] the mc was first studied for the direction exhibiting the maximum effect , namely the [ -554 ] crystal axis . figure [ mc111vsteparh](a ) shows the mc data collected with both @xmath4 and @xmath86 along [ -554 ] . by increasing t , the mc at the main peak ( `` point b '' ) is smoothly decreasing in magnitude and the position of the peak shifts to lower values of h@xmath28 . apart from the main peak , a small shoulder ( `` point a '' ) with some hysteresis
is observed at 700 oe , which can also be seen in the inset of fig .
[ mc111vsteparh](a ) .
the position of this shoulder is the same with the field - induced anomalies observed in the me@xmath7 and the magnetization measurements ( see inset of fig . [ mc4p2k ] ) .
the anomaly at `` point a '' is smeared out ( and thus it is not discernible ) between 20 - 30 k , but it can be clearly seen again above 30 k as a small dip and with opposite ( negative ) mc sign .
we recall here that such a different qualitative behavior above and below the region 20 - 30 k was also observed in the [ 100 ] magnetization measurements ( fig .
[ htdemcorrn2 ] ) , and thus the two effects must be correlated .
finally , the mc at the plateau is generally quite small and negative , but picks up a small positive value above 55 k , which however does not saturate up to 4 t. some of the above features can also be seen in fig .
[ mc111vsteparh](b ) , which shows the mc at `` point a '' , at `` point b '' , and at the onset of the 1/2-plateau ( rescaled by a factor of -6.6 ) .
it is noteworthy that the t - dependence of the mc at `` point b '' is very similar to the one at the position of the plateau . crystallographic orientation , while the magnetic field is perpendicular to the electric field .
( b ) the t - dependence of the magnitude of the mc at saturation and at the point `` b '' , taken from ( a).,scaledwidth=40.0% ] the mc was also measured with @xmath4 in the ( -554 ) plane . here
the ac - electric field @xmath86 is kept in the [ -554 ] direction , as shown in fig .
[ mc111vsteperph ] . as in fig .
[ mc111vsteparh ] , the mc is again dominated by a peak ( `` point b '' in fig .
[ mc111vsteperph ] ) slightly before the onset of the 1/2-plateau , where a negative mc is observed . at low fields ( 250 oe ) and for @xmath90 23 k , a step - like behavior is observed in the mc data ( `` point a '' ) .
the field position of this step coincides with the field - induced transition observed in our magnetization measurements ( not shown here ) .
this low - field step - like shape of the mc curves is smeared out for t@xmath50 23 k. the t - dependence of the mc at `` point b '' and at the position of the plateau is presented in fig .
[ mc111vsteperph](b ) .
the overall dependence is very similar to the data shown in fig .
[ mc111vsteparh ] . however , the mc at `` point b '' is lower for the in - plane configuration , while larger ( negative ) mc values are now observed at the plateau .
we have studied the angular dependence of the mc at 15 k for `` crystal b '' in two different configurations . in the first configuration shown in fig .
[ mc100vsphihor ] , @xmath86 is along [ 100 ] while @xmath4 rotates in the ( 100 ) plane .
both the amplitude and the shape of the mc curves change drastically upon changing the orientation of @xmath4 . as in the me@xmath7 experiments ( fig .
[ mevsphi100horiz ] ) , the maximum effect is observed when @xmath4 is parallel to the cube face diagonals . on the other hand ,
the minimum mc value is observed when @xmath4 is along the cube edges .
the mc at the position of the maximum ( `` point b '' ) is shown in fig .
[ mc100vsphihor](b ) as a function of the angle @xmath81 between @xmath4 and the [ 010 ] direction .
a sinusoidal angular dependence of the form @xmath91 is observed which suggests a free - energy term of the form @xmath92 which is quartic in the magnetic order parameter . as we explain in sec .
[ sec : theorymc ] , such a term arises naturally in the analytical expansion of the free - energy in powers of the me coupling mechanism that is related to the me@xmath7 effect . in this second configuration , @xmath86 is fixed along [ 100 ] while @xmath4 makes an out - of - plane rotation as shown in fig .
[ mcvsangle100katak ] .
the mc follows a sinusoidal angular dependence of the form @xmath93 , both at point b " and at the 1/2-plateau . as we discuss in sec .
[ sec : theorymc ] , this dependence suggests the presence of the more conventional @xmath94 free - energy term in the free energy . finally , it is quite interesting to contrast the large ( and positive ) mc measured e.g. at `` point b '' when @xmath95 ( first configuration ) to the vanishing mc when @xmath96 ( second configuration ) . and the magnetic field rotating on the ( 100 ) plane ( at @xmath80 , @xmath97 $ ] .
( b ) the angular dependence of the magnetocapacitance at point b " is given ( b).,width=325 ] , and @xmath4 rotating from [ 100 ] ( @xmath76 ) to [ 03 - 7 ] ( @xmath77 ) .
( b ) the angular dependence of the mc at the plateau and at point `` a '' , as taken from ( a).,width=325 ]
from the position of the characteristic anomalies observed in the me@xmath7 ( figs . [ newp111vsthpar111 ] and [ newp111vsthperp111 ] ) and mc ( figs . [ mc111vsteparh ] and
[ mc111vsteperph ] ) measurements we have deduced the t - h@xmath28 phase diagram ( fig . [ hcvstmcme ] ) of cu@xmath0oseo@xmath1 for both in and out - of - the ( -554 ) plane orientations of the magnetic field . for the out - of - plane configuration we have also included the data points from our magnetization measurements for comparison
. the field position at which the low field - induced transition and the plateau are probed from the dielectric measurements fits quite well with the corresponding positions from the magnetization measurements .
the main differences between fig .
[ hcvstmcme](b ) and fig .
[ hcvstmcme](a ) are due to the different demagnetizing field for the in - plane and the out - of - plane configuration .
altogether , the dielectric measurements are consistent with the physical picture we described in sec .
[ magnetization ] above , namely that we have at least two additional phases below the 1/2-plateau , the low - field ( @xmath98 ) and the intermediate - field ( @xmath99 ) phase .
the low - field phase is a multi - domain chiral phase whose me response defines two sub - regions ( not specified in fig .
[ hcvstmcme ] ) : the first shows a vanishing me@xmath7 effect which is consistent with the picture of multi - domain averaging of flat helices , while the second region marks a complex field - induced reorientation process which concludes at the `` point a '' peak anomalies .
the intermediate - field phase is related to the evolution of a single - domain conical helix under a magnetic field , with the propagation vector along the field .
in fact , the phenomenological theory presented in secs . [ sec : theoryme ] and [ sec : theorymc ] below , shall provide very strong arguments in favor of this picture .
in addition to the above phases , there are indications in our data for at least one more phase . for example ,
the second anomaly ( point a in fig . [ hcvstmcme](a ) ) observed in the small t - window ( 2k ) close to t@xmath44 might be related to the presence of the so - called `` a - phase '' that is found in similar compounds.@xcite " , b " , and the onset of the 1/2-plateau , as observed in the field - induced polarization ( fig .
[ newp111vsthpar111](a ) ) and mc measurements ( fig .
[ mc111vsteparh ] ( a ) ) .
( b ) same as in ( a ) , but the results are now from the polarization data of fig .
[ newp111vsthperp111](a ) and the mc data of fig .
[ mc111vsteperph](a).,scaledwidth=40.0% ]
here we develop a phenomenological landau - ginzburg approach to account for the me@xmath7 effect .
we shall focus on the intermediate - field and the 1/2-plateau phases . in this region ,
the magnetic state can be described by a single - domain conical helix with propagation vector @xmath101 along the field direction @xmath102 .
we shall keep track of both cu1 and cu2 magnetic order parameters by writing [ eqn : ops ] _
i^_h + _ i^ ( e^i _ i + h.c . ) with @xmath103 . here
@xmath104 , and @xmath105 form a triad with @xmath106 . since the cu1 and cu2 moments are antiparallel , @xmath107 and similarly @xmath108 .
the long - wavelength transverse twisting between two nearby moments @xmath109 and @xmath110 is captured by the relative phase factor @xmath111 ( where @xmath112 , with @xmath113 . [ cols="^,^,^,^,^",options="header " , ] we continue by recognizing that the magnetic and electric polarizations of the system form three - dimensional ( 3d ) representations of the present cubic group @xmath114 , but the two quantities transform differently under time - reversal .
table [ symmetries ] summarizes the transformation properties under the three point - group generators of @xmath114 plus the time - reversal invariance .
it is also useful for what follows to define the following composite 3d representations of p2@xmath193 & & \ { , } _ + = ( a^yb^z , a^zb^x , a^xb^y ) + & & \ { , } _ - = \ { , } _ + = ( a^zb^y , a^xb^z , a^yb^x ) given two vectors @xmath115 and @xmath116 . now , to leading order we may look at the invariant local terms that are linear in @xmath2 and quadratic in the magnetic order parameters . in particular , the theory may contain any pair among @xmath117 and @xmath118 . for any such pair ( @xmath109 and @xmath110 , in the following ) ,
there are only two invariants that do not involve spatial derivatives ( the latter are described in app .
[ app : me ] ) : & & _ i^y _ j^z p^x + _ i^z _ j^x p^y + _ i^x _ j^y p^z \ { _ i , _ j } _ + + & & _ i^z _ j^y p^x + _ i^x _ j^z p^y + _ i^y _ j^x p^z \{_j , _ i}_+ where @xmath119 are two composite 3d polar representations of @xmath114 , as defined above .
so to lowest order , the free energy reads [ eqn : fe ] ( \ { } , ) & = & _ 0(\ { } ) + - + & - & ( _ 1 \ { _ i , _ j } _ + + _ 2 \ { _ i , _ j } _ - ) .
here @xmath120 stands for the magnetic portion of the energy in the absence of the me coupling ( i.e. it contains the strong exchange energy plus the chiral dm term and other types of anisotropies ) , @xmath121 is the electric susceptibility ( scalar for cubic groups ) in the absence of me coupling , and @xmath122 are the me coupling constants . the physical origin of the last term of eq .
( [ eqn : fe ] ) becomes more transparent by rewriting it as , where @xmath123 .
the first term inside the brackets involves the symmetric exchange anisotropy , while the second involves the antisymmetric dm interaction .
these two anisotropy types show drastically different contributions to the me@xmath7 effect .
indeed , minimizing @xmath124 with respect to @xmath2 gives ( for @xmath125 ) : [ eqn : p1 ] /_e
= ( \{_i,_j } _
+ + \{_i,_j } _ - ) +
it is now evident that the me@xmath7 effect will be dominated by the symmetric exchange component , since nearby spins are almost collinear in the long - wavelength conical state .
thus , despite the fact that the dm coupling is the primary source of helimagnetism , its influence on the me@xmath7 effect is heavily diminished by its long - wavelength nature . to make these points more explicit and obtain the final predictions for the angular and field dependence of the me@xmath7 effect , we proceed by replacing eq .
( [ eqn : ops ] ) in eq .
( [ eqn : p1 ] ) one immediate prediction is that the helimagnetism induces both a uniform as well as a spatially oscillating polarization component which is also of mesoscopic nature . in the following we shall restrict ourselves to the uniform component , @xmath126 , which is the one we measure in our experiments . using the relation @xmath127 , and expanding @xmath128 , we find : [ eqn : p2 ] & = & _ + ( _ ^2-_^2 ) \ { _ h , _ h } _ + + & - & _
^2 ( _ ij ) _ h where @xmath129 is the sign of @xmath130 .
the second term is the contribution from the dm coupling and can be disregarded since @xmath131 , as discussed above .
similar contributions that scale linearly with @xmath132 arise also from the inhomogeneous lifshitz - type invariants in the free energy ( app .
[ app : me ] ) .
altogether we may write [ eqn : p3 ] & & _ + ( _ ^2 - _ ^2 ) \ { _ h , _ h } _ + which is our central expression . as we are going to show below , the angular dependence found in the experiments is fully captured by the projection of the vector @xmath133 along @xmath64 .
similarly , the observed field dependence can be fully captured ( in the field range where the magnetic state is described by eq .
( [ eqn : ops ] ) ) by the quantity @xmath134 .
this quantity goes from @xmath135 for a completely flat helix ( @xmath136 ) to + 1 at the 1/2-plateau ( where @xmath137 ) , and thus provides a physically transparent explanation for the sign difference between the measured polarization at `` point a '' and that at the 1/2-plateau .
of course , the `` point a '' does not correspond to a completely flat helix ( this would be the case at zero - field ) and this is why the relative magnitude @xmath138 is actually smaller than 1/2 . here
we comment what happens for a flat helix whose propagation vector is not along @xmath102 , but along some general direction @xmath140 .
this analysis explains why the total polarization averages out in the low - field region of the multi - domain helical phase .
given that @xmath141 is now equal to @xmath140 , one may easily show that the above formulas remain valid if we replace @xmath102 by @xmath140 everywhere .
in particular , the net polarization is given by ^ _
\{_k,_k}_+ .
now , the magnetization process close to the 1/2-plateau ( fig . [ mhloop5k](a ) ) suggests the @xmath142 as the easy axes , but the low - field data can not confirm this ( in fact they seem to suggest the @xmath143 as the easy axes but this is most likely related to the complex low - field domain reorientation process , as discussed above ) . in any case , one can show that the net polarization vanishes for both @xmath142 @xmath143 cases .
in the first case , @xmath144 , all three domains give a vanishing contribution , while in the second , @xmath145 , each of the four domains gives a finite contribution ( for a flat helix @xmath146 ) but the net polarization from all domains averages out .
we now proceed to a direct comparison to the experimental me@xmath7 data , beggining with the angular dependence . in the following
we shall use the scaled quantity = .
we begin with crystal a for which @xmath147 .
for the data shown in fig .
[ pvstheta111katak](b ) , @xmath148 , where @xmath149 .
applying eq .
( [ eqn : p2 ] ) we find [ eqn : akatak ] ^(-554 ) = - ( 55 + 150 ( 2 ) - 3 ( 2 ) ) the solid line shown in fig .
[ pvstheta111katak](b ) is a fit using this expression .
parenthetically , we should remark that the term proportional to @xmath150 in eq .
( [ eqn : akatak ] ) arises from the small deviation ( 6@xmath78 ) of @xmath64 from [ -111 ] .
this is why the proportionality constant in front of this term is much smaller than the first two terms of eq .
( [ eqn : akatak ] ) . would @xmath64 be exactly along [ -111 ] we would get ^(-111 ) = - .
these expressions explain in particular why the polarization data with @xmath151 ( @xmath76 ) are larger by a factor of -2 from the data corresponding to @xmath152 ( @xmath77 ) , see fig .
[ newp111vsthperp111 ] .
let us now turn to crystal b for which @xmath153 $ ] . for the data shown in fig .
[ mevsphi100horiz ] , the magnetic field rotates in the plane defined by @xmath154 + \sin\phi ~ [ 001]$ ] , and eq .
( [ eqn : p2 ] ) gives [ eqn : p100a ] ^(100 ) = ( 2 ) which is also in perfect agreement with the data .
similarly , for the data shown in fig .
[ pvstheta100katak ] , the magnetic field rotates in the plane defined by @xmath155 + \sin\theta ~ \vec{e}_b'$ ] , where @xmath156 , and eq .
( [ eqn : p2 ] ) gives [ eqn : p100b ] ^(100 ) = - ^2 , which is again in very good agreement with the data , apart from a very small shift which may originate either from a tiny misalignment of @xmath64 away from [ 100 ] , or from additional minor contributions to the me free energy .
let us now turn to the field - dependence of the measured polarization and compare it with eq .
( [ eqn : p2 ] ) . to this end
we shall make use of some well known and rather simple theoretical expressions for the evolution of @xmath157 and @xmath158 as a function of applied field , in the conical phase . following e.g. ref . , we write [ eqn : kirkpatrick ] _
= g , _ = g , where @xmath159 is a t - dependent prefactor which vanishes at @xmath31 , and @xmath54 is the onset field of the 1/2-plateau . for concreteness
we consider the data shown in fig .
[ mevsphi100horiz ] . according to eq .
( [ eqn : p100a ] ) , = ( 3 -1 ) ( 2 ) , where we note that the polarization scales quadratically with @xmath160 . from our data we have access to both parameters @xmath159 and @xmath54 ( the latter is influenced by the demagnetizing field ) , and so we can make a direct comparison between theory and experiment .
this comparison is shown in fig .
[ fig : p100exptheory ] for a few representative angles @xmath84 , and is remarkably successful .
the deviation around the 1/2-plateau can be ascribed to thermal fluctuations ( the experimental data are taken at 15 k while the above expression is valid only at t=0 ) .
altogether , this gives further confidence that the me coupling mechanism of eq .
( [ eqn : fe ] ) above captures the experimental findings both qualitatively and quantitatively .
for some representative angles @xmath84 ( data reproduced from fig .
[ mevsphi100horiz](a ) ) .
the shaded area shows the low - field phase ( which further consists of two sub - regions ) .
the first arrow designates the onset of the single - domain conical phase ( peak at `` point a '' ) for @xmath161 .
this phase evolves continuously toward the 3up-1down ferrimagnetic phase ( 1/2-plateau ) .
, scaledwidth=46.0% ]
to capture the physics of the mc , we must include higher order terms in the me portion of the free energy , and in particular we must look at quadratic terms in the polarization . to satisfy time - reversal invariance
this term must scale as @xmath162 , were @xmath163 is an even integer .
now the strong @xmath164 angular dependence of fig . [ mc100vsphihor](b ) suggests that we should take @xmath165 , while @xmath166 may also be present as suggested by the angular dependence of fig .
[ mcvsangle100katak](b ) .
the simplest possible term with @xmath165 , that is also physically motivated , is the square of the symmetric combination of the me@xmath7 invariant of eq .
( [ eqn : fe ] ) , namely [ eqn : mc ] _ 0=^2 = ( \{_1,_1}_+)^2 , where in the second equality we disregarded terms related to the small phase difference @xmath167 , as discussed above .
physically , @xmath168 stands for the second - order term in the analytical expansion of the free energy in the me coupling mechanism discussed in sec .
[ sec : theoryme ] . including this term gives [ eqn : femc1 ] = -= -_0 _ 0 , where @xmath169 is the coupling constant , and @xmath124 is given in eq .
( [ eqn : fe ] ) above .
the polarization is again obtained by minimizing with respect to @xmath2 .
however , in the experiment one uses an alternating ac - electric field and measures the alternating portion of the charge on the capacitance plates .
so from the derivative of @xmath170 with respect to @xmath2 we can disregard the terms that are related to the me@xmath7 effect .
this gives e^_ac = - _ 0 . with @xmath171 ,
we get ^-1 _ ^-1_-_e^-1 _ = -_0 .
now , given that the magnetocapacitance is a very small quantity ( mc @xmath172 1% ) , we may write @xmath173 , and thus @xmath174 .
if we are measuring the charge @xmath63 on a surface s ( which is perpendicular to the applied ac - electric field ) , then = = _ 0 _ e _ e_s^ e_s^. taking derivatives we find = 2 \{_1,_1}_+^ \{_1,_1}_+^for example , for @xmath175 ( which is directly relevant to most of the mc data we show here ) , @xmath176 . taking the uniform portion of @xmath177 for the conical helix
gives [ eqn : mc1 ] = f_1 ( e_h^y e_h^z)^2 + f_2 ( e_h^x)^2 + f_3 , where f_1(_a)&= & ( _ 1^)^4 - 3 ( _ 1^_1^)^2 + ( _ 1^)^4 [ eqn : f1 ] + f_2(_a)&= & ( _ 1^)^4 - ( _ 1^_1^)^2 + f_3(_a)&= & ( _ 1^_1^)^2 we now consider the data shown in fig .
[ mc100vsphihor](b ) for the angular dependence of the mc at `` point b '' , when the field rotates in the ( 100 ) crystal plane .
replacing @xmath178+\sin\phi[001]$ ] in eq .
( [ eqn : mc1 ] ) yields & = & f_1 ^2(2 ) + f_3 , which is exactly the angular dependence observed in fig .
[ mc100vsphihor](b ) .
this provides confidence that the invariant @xmath168 must be present in @xmath179 .
next , we consider the angular dependence shown in fig .
[ mcvsangle100katak](b ) . replacing @xmath180+\sin\phi[03 - 7]/\sqrt{58}$ ] in eq .
( [ eqn : mc1 ] ) now gives = f_1 ^4 + f_2 ^2 + f_3 . at the 1/2-plateau , @xmath181 , @xmath182 ( since @xmath183 ) , and
so this expression predicts that @xmath184 , which is clearly in disagreement with the @xmath185 behavior of the data .
so we may conclude that there is at least one more invariant giving a @xmath185 contribution .
in particular , this term must scale quadratically with the magnetic order parameter ( @xmath166 ) to account for the angular dependence . a list of such invariant terms is provided and discussed in app .
[ app : mc ] .
we have observed the me effect in single crystals of the cubic compound cu@xmath0oseo@xmath1 , and we have studied its angular and temperature dependence .
we have demonstrated that the electric polarization can be controlled by an external magnetic field .
we have also investigated the temperature , field and angular dependence of the magnetocapacitance .
both the magnetic field - induced polarization p and the magnetocapacitance mc set in at the magnetic ordering temperature and show a number of anomalous features which we ascribe to magnetic phase transitions . from the positions of these anomalies we have mapped out the phase diagram of the system in the @xmath5-t plane . in the field region below the 3up-1down collinear ferrimagnetic phase and for @xmath18630 k we find two additional magnetic phases , whose origin is related to the presence of a dzyalozinskii - moriya chiral term in the free energy which impairs the homogeneity of the magnetic ordering and
gives rise to a continuous helical twisting of the ferrimagnetic order parameter with very long wavelength .
the vanishing of the me response at the low - field phase of cu@xmath0oseo@xmath1 suggests a multi - domain structure of flat helices , where the propagation vectors of the different helices are pinned along the preferred axes of the system . in the intermediate - field phase
the propagation vectors of the helices are aligned along the applied magnetic field @xmath4 . for t@xmath26 30
k our magnetization measurements suggest a more complicated state and a possible thermal reorientation of the anisotropy . furthermore we also find a narrow range ( 2k ) close to the transition temperature @xmath31 with a double - peak anomaly in the me@xmath7 effect which might be due to the possible presence of the `` a - phase '' . we have also developed a landau - ginzburg phenomenological theory which accounts for the me@xmath7 effect in the intermediate - field and the 1/2-plateau phase , and explains the angular , temperature and field dependence of this effect in a simple and physically transparent way .
contrary to expectations , the influence of the dm coupling in the me effect is heavily diminished due to the long - wavelength nature of the twist . instead , the me effect is driven by an exchange striction mechanism involving the symmetric portion of the exchange anisotropy , with three central features : ( i ) the uniform component of @xmath2 points along the vector @xmath8 ; ( ii ) its strength is proportional to @xmath9 , where @xmath10 is the longitudinal and @xmath11 is the transverse ( and spiraling ) component of the magnetic ordering .
hence , the field dependence of p provides a clear signature of the evolution of a conical helix under a magnetic field ; and ( iii ) apart from its uniform component , the polarization @xmath2 has also a long - wavelength spatial component that is given by the pitch of the magnetic helices .
this effectively mesoscale antiferroelectric structure could in principle be probed by experiments with high spatial resolution , which may further clarify the me coupling mechanism in this cu@xmath0oseo@xmath1 at a more microscopic level .
we acknowledge experimental assistance from k. schenk , and c. amendola .
we would also like to thank a.n .
bogdanov , o. janson , h. rosner , k. schenk , m. wolf , and s. wurmehl for useful discussions .
one of the authors ( h. b. ) acknowledges financial support from the swiss nsf and by the nccr manep .
here we analyze the lifshitz - type invariants between magnetic order parameters and dielectric polarization.@xcite these ( free ) energy terms are linear in one spatial derivative of the magnetic order parameter and are allowed in the present compound . as we explained above and show also below , these me terms are most likely not related to the observed me@xmath7 effect , despite the fact that they give the same angular dependence with the term of eq .
( [ eqn : fe ] ) . however , in a systematic expansion of the me free energy they should not be omitted , as they yield terms linear in @xmath187 and may modify certain aspects of the me behavior , in particular in systems with shorter pitch magnetic helices . the lifshitz invariants are the following seven terms & & l_1 = \{,_i}_+ \{,_j}_+ , l_2 = \{,_i}_+ \{,_j}_- + & & l_3 = \{,_i}_- \{,_j}_+ , l_4 = \{,_i}_- \{,_j}_- + & & l_5 = p^x _ i^x ^x _ j^x + p^y _ i^y ^y _ j^y + p^z _
i^z ^z _ j^z + & & l_6 = p^x _ i^x ^y _ j^y + p^z _
i^z ^x _ j^x + p^y _
i^y ^z _ j^z + & & l_7 = p^x _ i^x ^z _ j^z + p^z _
i^z ^y _ j^y + p^y _
i^y ^x _ j^x and their contribution to the free energy is _
l = -_=1 ^ 7 _ l _ , where @xmath188 are the corresponding coupling constants .
the contribution @xmath189 of these terms to the polarization can be found again by differentiating with respect to @xmath2 .
for example , for @xmath125 : & = & _ 1 _ i^y ^x _ j^y + _ 2 _ i^y ^y _ j^x + _ 3 _ i^z ^z _
j^x + _ 4 _
j^z + & + & _ 5 _ i^x ^x _ j^x + _ 6 _
i^x ^y _ j^y + _ 7 _ i^x ^z _ j^z .in order to account properly for the spatial derivatives , we may rewrite the position of a given cu site as @xmath190 , where @xmath191 denotes the relative position of that site with respect to the center of the corresponding unit cell @xmath192 . now , from each term @xmath193 there will be two types of contributions , one proportional to @xmath194 and another proportional to @xmath195 .
the latter will be disregarded in the following since it is quadratic in @xmath132 . with this in mind , and keeping only the uniform component ( _ i^^_j^)^ & = & ( k _ i^_j^ ) e_h^ _ _ e_h^one can easily show that only @xmath196 and @xmath197 contribute to the uniform portion of @xmath198 , and the same happens for @xmath199 and @xmath200 . the final expression for @xmath201 is _
l = _ e ( _ 3-_2 ) k _
i^_j^ \ { _ h,_h } _ + , which has exactly the same angular dependence as in eq .
( [ eqn : p2 ] ) above .
the difference is that here the pre - factor scales with @xmath202 which can not account for the observed field dependence , and in particular the sign change of @xmath2 between the `` point a '' and the 1/2-plateau ( see sec . [
sec : theoryme ] ) .
here we provide a list of invariant terms that are quadratic in p and thus may contribute to the mc , in addition to the term @xmath168 that we discussed in sec .
[ sec : theorymc ] .
these invariants are quadratic ( in contrast to @xmath168 which is quartic ) in the magnetic order parameters and as such , they provide the @xmath185 angular dependence that is needed for the full interpretation of the data shown in fig . [
mcvsangle100katak](b ) .
we restrict ourselves to spatially uniform invariants
. these are the following five _
1&=&\ { _ i,_j } _ + \ { , } _ + + _ 2&=&\ { _ i,_j } _ - \ { , } _ + + _ 3&=&p_x^2 _ i^x_j^x + p_z^2 _ i^z_j^z+p_y^2 _ i^y_j^y + _ 4&=&p_x^2 _ i^y_j^y + p_z^2 _ i^x_j^x+p_y^2 _ i^z_j^z + _ 5&=&p_x^2 _ i^z_j^z + p_z^2 _ i^y_j^y+p_y^2 _ i^x_j^x including these invariants in eq .
( [ eqn : femc1 ] ) gives [ eqn : femc2 ] = -= -_0 _ 0 - _ = 1 ^ 5 _ _ where @xmath203 are the new coupling constants . repeating the steps described in sec .
[ sec : theorymc ] leads to = -2 _ 0 ( _ 1^y _ 1^z ) ^2 -2 ( _ 3 _ i^x_j^x + _ 4 _
i^y_j^y + _ 5 _ i^z_j^z ) , where the second term gives the contributions from the invariants @xmath204-@xmath205 . using ( _ 1^_1^)^&=&e_h^e_h^+ _ ( _ 1^)^2
, we find that the contribution to the mc from @xmath204-@xmath205 is & = & + & + & ( _ 3+_4+_5 ) ( _ 1^)^2 . where @xmath206 is again the sign of @xmath207 .
99 p. curie , j. phys .
3rd series , * 3 * , 393 , ( 1894 ) .
l. d. landau , and e. m. lifshitz , electrodynamics of continuous media ( pergamon press , oxford , 1984 ) .
h. katsura , n. nagaosa , a. v. balatsky , * 95 * , 057205 , ( 2005 ) .
m. mostovoy , * 96 * , 067601 , ( 2006 ) .
i. e. dzyaloshinskii , sov .
jetp * 10 * , 628 ( 1960 ) .
d. n. astrov , sov .
jetp * 11 * , 708 ( 1960 ) .
g. t. rado and v j. folen , * 7 * , 310 , ( 1961 ) .
k. f. wang , j .-
liu , and z. f. ren , advances in physics * 58 * , 321 ( 2009 ) .
n. a. hill , j. phys .
b * 104 * , 6694 ( 2000 ) ; annu . rev .
32 * , 1 ( 2002 ) .
d. khomskii , physics * 2 * , 20 ( 2009 ) .
w. eerenstein , n. d. mathur , and j. f. scott , nature * 442 * , 759 ( 2006 ) .
n. a. spaldin , m. fiebig , science * 309 * , 391 ( 2005 ) .
m. fiebig , j. phys .
d * 38 * , r123 ( 2005 ) .
m. gajek , m. bibes , s. fusil , k. bouzehouane , j. fontcuberta , a. barthlmy , and a. fert , nature materials * 6 * , 296 ( 2007 ) .
v. e. wood and a. e. austin , int
. j. magn . * 5 * , 303 ( 1974 ) .
g. a. smolenski and i. e. chupis , sov .
* 25 * , 475 ( 1982 ) .
h. schmid , ferroelectrics * 162 * , 317 ( 1994 ) .
t. kimura , s. kawamoto , i. yamada , m. azuma , m. takano , and y. tokura , * 67 * , 180401 ( 2003 ) .
j. wang , j. b. neaton , h. zheng , v. nagarajan , s. b. ogale , b. liu , d. viehland , v. vaithyanathan , d. g. schlom , u. v. waghmare , n. a. spaldin , k. m. rabe , m. wuttig , and r. ramesh , science * 299 * , 1719 ( 2003 ) . t. kimura , t. goto , h. shintani , k. ishizaka , t. arima , and y. tokura , nature * 426 * , 55 ( 2003 ) . n. hur , s. park , p. a. sharma , j. s. ahn , s. guha , and s .- w .
cheong , nature * 429 * , 392 ( 2004 ) .
g. lawes , a. b. harris , t. kimura , n. rogado , r. j. cava , a. aharony , o. entin - wohlman , t. yildrim , m. kenzelmann , c. broholm , and a. p. ramirez , * 95 * , 087205 ( 2005 ) . g. lautenschlger , h. weitzel , t. vogt , r. hock , a. bhm , m. bonnet , and h. fuess , * 48 * , 6087 ( 1993 ) . k. taniguchi , n. abe , t. takenobu , y. iwasa , and t. arima , * 97 * , 097203 ( 2006 ) .
y. j. choi , h.
t. yi , s. lee , q. huang , v. kiryukhin , and s .- w .
cheong , * 100 * , 047601 ( 2008 ) . i. a. sergienko and e. dagotto , * 73 * , 094434 ( 2006 ) .
li , s. dong , and j .-
liu , * 77 * , 054442 ( 2008 ) .
j. hu , * 100 * , 077202 ( 2008 ) . v. g. baryakhtar , v.a .
lvov , d. a. yablonskii , jetp lett . * 37 * , 673 ( 1983 ) .
stefanovskii , ferroelectrics * 161 * , 245 ( 1994 ) . j. j. betouras , g. giovannetti , j. van den brink , * 98 * , 257602 ( 2007 ) .
h. effenberger , and f. pertlik , monatsh . chem .
* 117 * , 887 , ( 1986 ) .
g. bos , c. v. colin , and t. t. m. palstra , * 78 * , 094416 , ( 2008 ) .
a. larraaga , j. l. mesa , l. lezama , j. l. pizarro , m. i. arriortua , and t. rojo , materials research bulletin , * 44 * , 1 , ( 2009 ) .
v. p. gnezdilov , k. v. lamonova , yu .
g. pashkevich , p. lemmens , h. berger , f. bussy , and s. l gnatchenko , fizika nizkikh temperatur * 36 * , 688 , ( 2010 ) .
m. i. kobets , k. g. dergachev , e. n. khatsko , a. i. rykova , p. lemmens , d. wulferding , h. berger , low temp .
phys . * 36 * , 176 ( 2010 ) m. belesi , i. rousochatzakis , h. c. wu , h. berger , i. v. shvets , f. mila , j. p. ansermet , * 82 * , 094422 , ( 2010 ) . k. h. miller , x. s. xu , h. berger , e. s. knowles , d. j. arenas , m. w. meisel , and d. b. tanner , * 82 * , 144107 , ( 2010 ) .
m. belesi , t. philippe , i. rousochatzakis , h. c. wu , h. berger , s. granville , i. v. shvets , j. -ph .
ansermet , journal of physics : conference series * 303 * , 012069 ( 2011 ) . c. l. huang , k. f. and tseng , c. c. chou , s. mukherjee , j. l. her , y. h. matsuda , k. kindo , h. berger , h. d. yang , * 83 * , 052402 , ( 2011 ) .
a. maisuradze , z. guguchia , b. graneli , h. m. rnnow , h. berger , h. and h. keller , phys .
b * 84 * , 064433 ( 2011 ) ; a. maisuradze , a. shengelaya , h. berger , d. m. djoki , and h. keller , arxiv:1204.2481v1 s. seki , x. z. yu , s. ishiwata , and y. tokura , science * 336 * , 198 ( 2012 ) . i. e. dzyaloshinskii , sov . phys .
jetp * 19 * , 960 ( 1964 ) ; * 20 * , 223 ( 1965 ) ; * 20 * , 665 ( 1965 ) .
i. e. dzyaloshinskii , sov .
jetp * 5 * , 1259 ( 1957 ) .
t. moriya , phys . rev . * 120 * , 91 ( 1960 ) .
p. bak and m. h. jensen , j. phys .
c : solid state phys . * 13 * , l881 ( 1980 ) ; o. nakanishi , a. yanase , a. hasegawa , m. kataoka , solid state . comm . * 35 * , 995 ( 1980 ) .
y. ishikawa , k. tajima , d. bloch , m. roth , solid state commun .
* 19 * , 525 ( 1976 ) .
j. beille , j. voiron , m. roth , solid state commun . * 47 * , 399 ( 1983 ) .
b. lebecht , j. bernhard , t. freltoft , j. phys . : condens .
matter * 1 * , 6105 ( 1989 ) .
a. n. bogdanov and d. a. yablonskii , sov .
jetp * 68 * , 101 ( 1989 ) ; a.n.bogdanov and a.hubert , j. magn .
magn . mater . *
138 * , 255 ( 1994 ) .
a. n. bogdanov , u. k. rler , c. pfleiderer , physica b * 359 - 361 * , 1162 ( 2005 ) .
u. k. rler , a. n. bogdanov , c. pfleiderer , nature * 442 * , 797 ( 2006 ) .
x. z. yu , y. onose , n. kanazawa , j. h. park , j. h. han , y. matsui , n. nagaosa , y. tokura , nature * 465 * , 901 ( 2010 ) .
x. z. yu , n. kanazawa , y. onose , k. kimoto , w. z. zhang , s. ishiwata , y. matsui , y. tokura , nature mater .
* 10 * , 106 ( 2011 ) .
h. wilhelm , m. baenitz , m. schmidt , u. k. rler , a. a. leonov , a. n. bogdanov , * 107 * , 127203 ( 2011 ) .
u. k. rler , a. a. leonov , a. n. bogdanov , j. phys .
* 303 * , 012105 ( 2011 ) .
h. wilhelm , m. baenitz , m. schmidt , c. naylor , r. lortz , u. k. rler , a. a. leonov , a. n. bogdanov , to appear in j. phys . : condens .
matter ( 2012 ) ; arxiv:1203.4964 s. kusaka , k. yamamoto , t. komatsubara , y. ishikawa , solid state commun . *
20 * , 925 ( 1976 ) .
t. komatsubara , s. kusaka , and y. ishikawa , in _
proceedings 6@xmath208 intern .
internal friction and ultrasonic attenuation in solids _
( university of tokyo press , 1977 ) p. 237 .
k. kadowaki , k. okuda , m. date , j. phys .
jpn , * 51 * , 2433 ( 1982 ) . y.ishikawa and m.arai , j. phys .
. jpn . * 53 * , 2726 ( 1984 ) . c. i. gregory , d. b. lambrick , n. r. bernhoeft , j. magn .
magn . mater . *
104 - 107 * , 689 ( 1992 ) .
s. endo , h. matsuzaki , f. ono , t. kanomata , t. kaneko , j. magn .
magn . mater . *
140 * , 119 ( 1995 ) . c. thessieu , c. pfleiderer , a. n. stepanov , j. flouquet , j. phys .
matter * 9 * , 6677 ( 1997 ) .
d. lamago , r. georgii , c. pfleiderer , p. bni , physica b * 385 - 386 * , 385 ( 2006 ) . s. v. grigoriev , s. v. maleyev , a. i. okorokov , yu .
o. chetverikov , h. eckerlebe , * 73 * , 224440 ( 2006 ) .
s. v. grigoriev , s. v. maleyev , a. i. okorokov , yu .
o. chetverikov , p. bni , r. georgii , d. lamago , h. eckerlebe , k. pranzas , * 74 * , 214414 ( 2006 ) . s. m hlbauer , b. binz , f. jonietz , c. pfleiderer , a. rosch , a. neubauer , r. georgii , p. bni , science , * 323 * , 915 ( 2009 ) . a. neubauer , c. pfleiderer , b. binz , a. rosch , r. ritz , p. g. niklowitz , p. bni , * 102 * , 186602 ( 2009 ) .
a. bauer , a. neubauer , c. franz , w. mnzer , m. garst , c. pfleiderer , * 82 * , 064404 ( 2010 ) . s. m. stishov , a. e. petrova , s. khasanov , g. kh .
panova , a. a. shikov , j. c. lashley , d. wu , t. a. lograsso , j. phys . : condens .
matter * 20 * , 235222 ( 2008 ) .
a. e. petrova and s. m. stishov , j. phys . : condens .
matter * 21 * , 196001 ( 2009 ) . c. pappas , e. lelivre - berna , p. falus , p. m. bentley , e. moskvin , s. grigoriev , p. fouquet , b. farago , * 102 * , 197202 ( 2009 ) . c. pappas , e. lelivre - berna , p. bentley , p. falus , p. fouquet , b. farago , * 83 * , 224405 ( 2011 ) . c. pfleiderer , d. reznik , l. pintschovius , h. v. lhneysen , m. garst , a. rosch , nature * 427 * , 227 ( 2004 ) .
p. pedrazzini , h. wilhelm , d. jaccard , t. jarlborg , m. schmidt , m. hanfland , l. akselrud , h. q. yuan , u. schwarz , yu .
grin , f. steglich , * 98 * , 047204 ( 2007 ) .
s. v. grigoriev , s. v. maleyev , e. v. moskvin , v. a. dyadkin , p. fouquet , h. eckerlebe , * 81 * , 144413 ( 2010 ) . s. v. grigoriev , e. v. moskvin , v. a. dyadkin , d. lamago , t. wolf , h. eckerlebe , s. v. maleyev , * 83 * , 224411 ( 2011 ) . o. janson , a. a. tsirlin , h. rosner , dpg spring meeting abstract 2011 ( unpublished ) . j .-
rivera , ferroelectrics * 161 * , 165 ( 1994 ) .
rivera , eur .
j. b * 71 * , 299 ( 2009 ) .
r. i. joseph and e. schlmann , j. appl . phys . * 36 * , 1579 ( 1965 ) ; a. aharoni , ibid .
83 , 3432 ( 1998 ) .
m. l. plumer , m. b. walker , j. phys .
c * 14 * , 4689 ( 1981 ) .
k .- y . ho , t. r. kirkpatrick , y. sang , and d. belitz , * 82 * , 134427 ( 2010 ) . | we present magnetodielectric measurements in single crystals of the cubic spin-1/2 compound cu@xmath0oseo@xmath1 . a magnetic field - induced electric polarization
( @xmath2 ) and a finite magnetocapacitance ( mc ) is observed at the onset of the magnetically ordered state ( @xmath3 k ) .
both @xmath2 and mc are explored in considerable detail as a function of temperature ( t ) , applied field @xmath4 , and relative field orientations with respect to the crystallographic axes .
the magnetodielectric data show a number of anomalies which signal magnetic phase transitions , and allow to map out the phase diagram of the system in the @xmath5-t plane . below the 3up-1down collinear ferrimagnetic phase , we find two additional magnetic phases .
we demonstrate that these are related to the field - driven evolution of a long - period helical phase , which is stabilized by the chiral dzyalozinskii - moriya term @xmath6 that is present in this non - centrosymmetric compound .
we also present a phenomenological landau - ginzburg theory for the me@xmath7 effect , which is in excellent agreement with experimental data , and shows three novel features : ( i ) the polarization @xmath2 has a uniform as well as a long - wavelength spatial component that is given by the pitch of the magnetic helices , ( ii ) the uniform component of @xmath2 points along the vector @xmath8 , and ( iii ) its strength is proportional to @xmath9 , where @xmath10 is the longitudinal and @xmath11 is the transverse ( and spiraling ) component of the magnetic ordering .
hence , the field dependence of p provides a clear signature of the evolution of a conical helix under a magnetic field .
a similar phenomenological theory is discussed for the mc . |
understanding the dynamic phenomena in classical and quantum many - body systems is of importance for the most of areas of physics and chemistry .
the development of efficient algorithms for solving the equations of motion in such systems should therefore impact a lot of fields of fundamental research . during the last decade
a considerable activity @xcite has been directed on the construction of symplectic time - reversible algorithms that employ decompositions of the evolution operators into analytically solvable parts .
the decomposition algorithms exactly preserve all poincar invariants and , thus , are ideal for long - time integration in molecular dynamics @xcite and astrophysical @xcite simulations .
the reason is that for these algorithms the errors in energy conservation appear to be bounded even for relatively large values of the size of the time step .
this is in a sharp contrast to traditional runge - kutta and predictor - corrector schemes , where the numerical uncertainties increase linearly with increasing the integration time @xcite .
the main attention in previous investigations has been devoted to derive different - order decomposition algorithms involving only force evaluations during the time propagation .
for instance , the widely used velocity- and position - verlet algorithms @xcite relate , in the general classification , to a three - stages decomposition scheme of the second order with one force evaluation per step .
the fourth - order algorithm by forest and ruth @xcite corresponds to a scheme with three such force recalculations and consists of seven single - exponential stages .
sixth - order schemes are reproduced @xcite beginning from fifteen stages and seven evaluations of force for each body in the system per given time step . with
further increasing the order of force decomposition schemes , the number of stages and thus the number of the corresponding non - linear equations ( which are necessary to solve numerically to obtain the required time coefficients for single - exponential propagations ) increases drastically .
in addition , such equations become too cumbersome and all these , taking into account the capabilities of modern supercomputers , led to the impossibility of representing the direct decomposition algorithms of order eighth and higher in an explicit form @xcite . in order to simplify this problem ,
it was proposed to derive higher - order integrators by composing schemes of lower ( actually second ) orders .
the resulting second - order - based composition algorithms have been explicitly obtained up to the tenth order @xcite .
relatively recently @xcite , a deeper analysis of the operator factorization process has shown that the class of analytically integrable decomposition integrators can be extended including additionally a higher - order commutator into the single - exponential propagations . as a consequence ,
a set of new so - called force - gradient algorithms of the fourth order has been introduced .
a distinguishable feature of these algorithms is the possibility to generate solutions using only positive values for time coefficients during each substage of the integration .
this is contrary to the original decomposition approach , where beyond second order ( as has been rigorously proved by suzuki @xcite ) any scheme expressed in terms of only force evaluation must produce some negative time coefficients .
we mention that applying negative time propagations is impossible , in principle , in such important fields as non - equilibrium statistical mechanics , quantum statistics , stochastic dynamics , etc . , because one can not simulate diffusion or stochastic processes backward in time nor sample configurations with negative temperatures . in the case of stochastic dynamics simulations
it has been demonstrated explicitly @xcite that using fourth - order force - gradient algorithms leads to much superior propagation over standard verlet - based schemes of the second order in that it allows much larger time steps with no loss of precision .
a similar pattern was observed in classical dynamics simulations comparing the usual fourth - order algorithm by forest and ruth with its force - gradient counterparts @xcite .
quite recently , chin and kidwell @xcite has considered a question of how to iterate the force - gradient algorithms to higher order .
the iteration was based on creutz s and gocksch s approach @xcite according to which an algorithm of order @xmath1 can be obtained by triplet construction of a self - adjoint ( i.e. time - reversible ) scheme of order @xmath2 . then starting from a fourth - order integrator
, it has been shown in actual celestial mechanics simulations that for orders 6 , 8 , 10 , and 12 , the numerical errors corresponding to the force - gradient - based schemes are significantly smaller than those of the schemes basing on iterations of usual non - gradient algorithms .
the resulting efficiency of the integration has also increased considerably despite an increased computational efforts spent on the calculations of force gradients .
the same has been seen in the case of quantum mechanics simulations when solving the time - dependent schrdinger equation @xcite .
it is worth emphasizing , however , that the iteration scheme by chin and kidwell is far to be optimal for deriving high - order integrators belonging to the force - gradient class .
the reason is that the number of total force and its gradient evaluations increases too rapidly with increasing @xmath2 .
remembering that such evaluations constitute the most time - consuming part of the calculations , this may restrict the region of applicability of force - gradient algorithms to relative low orders only .
note that high - order computations are especially desirable in problems of astrophysical interest , because than one can observe over a system during very long times .
they may also be useful in highly precise molecular dynamics and quantum mechanics simulations to identify or confirm very subtle effects . in the present paper
we propose a general approach to construction of symplectic force - gradient algorithms of arbitrary orders .
the approach considers the splitting and composing of the evolution operators on the basic level , taking into account the explicit structure of truncation terms at each given order in the time step .
this has allowed us to obtain exclusively precise and economical algorithms with using significantly smaller number of single - exponential propagations than that appearing within standard decomposition and iteration schemes .
the paper is organized as follows .
the equations of motion for classical and quantum systems are presented in section ii.a .
the integration of these equations by direct decompositions and their force - gradient generalization are described in section ii.b .
explicit expressions for basic force - gradient algorithms of orders 2 , 4 , 6 , and 8 are also given there .
the higher - order integration basing on advanced compositions of lower - order schemes is considered in section ii.c .
the composition constants for fourth- , sixth- , and eighth - order - based schemes are calculated and written down in the same section up to the overall order 16 . sections iii.a and iii.b are devoted to applications of obtained force - gradient algorithms to molecular dynamics and celestial mechanics simulations , respectively . a comparative analysis of the new algorithms with existing integrators
is made there as well . the final discussion and concluding remarks are highlighted at the end in section iv .
consider first a classical @xmath3-body system described by the hamiltonian @xmath4 where @xmath5 is the position of particle @xmath6 moving with velocity @xmath7 and carrying mass @xmath8 , @xmath9 denotes the interparticle potential of interaction , and @xmath10 and @xmath11 relate to the total kinetic and potential energies , respectively . then the equations of motion can be presented in the following compact form @xmath12 \equiv l { \mbox{\boldmath $ \rho$}}(t ) \ , .\ ] ] here @xmath13 is the full set ( @xmath14 ) of phase variables , @xmath15 $ ] represents the poisson bracket and @xmath16 is the liouville operator with @xmath17 being the force acting on particles due to the interactions . in the case of quantum systems ,
the state evolution can be described by the time - dependent schrdinger equation @xmath18 where @xmath19 and @xmath20 are the kinetic and potential energy operators , respectively , and @xmath21 is the wave function .
so - called quantum - classical dynamics models @xcite can also be introduced .
this leads to a coupled system of newtonian ( 2 ) and schrdinger ( 4 ) equations .
but , in order to simplify notations , we restrict ourselves to the above purely classic and quantum considerations .
if an initial configuration @xmath22 or @xmath23 is provided , the unique solution to eq .
( 2 ) or ( 4 ) can be formally cast as @xmath24 where @xmath25 and @xmath26 are the size of the single time step and the total number of steps , respectively , @xmath27 denotes either @xmath28 or @xmath21 , whereas @xmath29 corresponds to @xmath30 or @xmath31 . as is well known , the time evolution of many - particle systems can not be performed exactly in the general case .
thus , the problem arises on evaluating the propagator @xmath32 by numerical methods .
the main idea of decomposition integration consists in factorization of the full exponential operator @xmath33 on such subpropagators which allow to be evaluated analytically or at least be presented in quadratures . within the original approach ,
this is achieved by splitting the operator @xmath34 into its kinetic @xmath35 and potential @xmath36 parts , where @xmath37 or @xmath38 and @xmath39 with @xmath40 being the acceleration or @xmath41 for the cases of classical or quantum mechanics , respectively .
then , taking into account the smallness of @xmath25 , the total propagator can be decomposed @xcite using the formula @xmath42 where the coefficients @xmath43 and @xmath44 are chosen in such a way to provide the highest possible value for @xmath45 at a given integer number @xmath46 . as a result ,
integration ( 5 ) can performed approximately with the help of eq .
( 6 ) by neglecting truncation terms @xmath47 .
the precision will increase with increasing the order @xmath2 and decreasing the size @xmath25 of the time step .
as can be verified readily , the exponential subpropagators @xmath48 and @xmath49 , appearing in the right - hand - side of eq . ( 6 ) , are analytically integrable for classical systems . indeed , taking into account the independence of @xmath50 on @xmath51 and @xmath52 on @xmath50 yields @xmath53 \\ [ -5pt ] & & { \rm e}^{{\cal b } \tau } { \mbox{\boldmath $ \rho$ } } \equiv { \rm e}^{\tau { \bf a } { \mbox{\boldmath $ \cdot$ } } \partial/\partial { \bf v } } \ { { \bf r } , { \bf v } \ } = \ { { \bf r } , { \bf v } + { \bf a } \tau \ } \nonumber\end{aligned}\ ] ] that represent simple shift operators in position and velocity spaces , respectively , with @xmath54 being equal to @xmath55 or @xmath56 . for quantum mechanics
propagations , the kinetic part @xmath57 will require carrying out two , one direct and one inverse , spatial fourier transforms @xcite , whereas the calculation of @xmath58 is trivial . in view of decompositions ( 6 ) , one can reproduce integrators of various orders in the time step . in particular , the well - known second - order ( @xmath59 )
velocity - verlet algorithm @xcite @xmath60 is readily derived from eq .
( 6 ) by putting @xmath61 with @xmath62 , @xmath63 , and @xmath64 .
the fourth - order ( @xmath65 ) algorithm by forest and ruth @xcite is obtained from eq .
( 6 ) at @xmath66 with @xmath62 , @xmath67 , @xmath68 , @xmath69 and @xmath70 , where @xmath71{2})$ ] .
schemes of the sixth order ( @xmath72 ) are derivable starting from @xmath73 with numerical representation of time coefficients @xcite .
the original decomposition approach has , however , a set of disadvantages .
first of all , it is worth pointing out that with further increasing the order of integration ( 6 ) to @xmath74 and higher , the number @xmath75 of unknowns @xmath43 and @xmath44 begins to increase too rapidly .
this leads to the impossibility of representing algorithms of such a type for @xmath76 in an explicit form @xcite , because it becomes impossible to solve the same number of the resulting cumbersome non - linear equations ( with respect to @xmath43 and @xmath44 ) even using the capabilities of modern supercomputers .
another drawback consists in the fact that for @xmath77 it is impossible @xcite at any @xmath78 to derive from eq .
( 6 ) a decomposition scheme with the help of only positive time coefficients .
for example , in the case of forest - ruth integration , three of eight coefficients , namely , @xmath79 , @xmath80 , and @xmath81 , are negative . as was mentioned in the introduction , schemes with negative time coefficients have a restricted region of application and are not acceptable for simulating non - equilibrium , quantum statistics , stochastic and other important processes . moreover , for schemes expressed in terms of force evaluation only , the main term @xmath82 of truncation uncertainties appears to be , as a rule , too big , resulting in decreasing the efficiency of the computations
. from the afore said , it is quite desirable to introduce a more general approach which is free of the above disadvantages . at the same time , this approach , like the original scheme , must be explicit , i.e. , lead to analytical propagations .
in addition , it is expected that the already known decomposition algorithms should appear from it as particular cases .
let us first analyze the structure of third - order truncation errors @xmath83 of the velocity - verlet algorithm in detail .
expanding both the sides of eq .
( 8) into taylor
s series with respect to @xmath25 , one finds @xmath84 + \frac{1}{24 } [ { \cal b},[{\cal a},{\cal b } ] ] \right ) \delta
t^3 + { \cal o}(\delta t^5)\ ] ] where @xmath85 $ ] denotes the commutator of two operators . taking into account the explicit expressions for operators @xmath35 and
@xmath36 it can be shown that one of the two third - order operators in eq .
( 9 ) , namely @xmath86 $ ] , is relatively simple and , that is more important , it allows to be handled explicitly , contrary to the operator @xmath87 $ ] . in the case of classical systems
it can be obtained readily that @xmath88 = \sum_{i=1}^n \frac{{\bf g}_i}{m_i } { \mbox{\boldmath $ \cdot$ } } \frac{\partial}{\partial { \bf v}_i } \equiv { \bf g } { \mbox{\boldmath $ \cdot$ } } \frac{\partial}{\partial { \bf v } } \ , , \ ] ] where @xmath89 . in view of the expression @xmath90 for forces ,
the required force - gradient evaluations @xmath91 are explicitly representable , i.e. , @xmath92 \\ [ -9pt ] \times \big({\bf r}_{ij } { \mbox{\boldmath $ \cdot$ } } ( { \bf a}_i-{\bf a}_j ) \big ) \bigg ] \equiv \sum_{j ( j \ne i)}^n { \bf g}({\bf r}_{ij } ) = { \bf g}_i({\bf r } ) \ , .
\nonumber\end{aligned}\ ] ] as can be seen easily from eqs .
( 10 ) and ( 11 ) , the operator @xmath93 commutes with @xmath94 , and , in addition , the function @xmath95 like @xmath96 does not depend on velocity
. then the force - gradient part @xmath97 of truncation uncertainties ( 9 ) can be extracted by transferring them from the left - hand - side of eq .
( 8) to its right side and further symmetrically collecting with operator @xmath36 under exponentials .
this yields the following force - gradient version @xmath98 of the velocity - verlet integrator , where already @xmath99 \delta t^3/12 $ ] . in the case of higher - order ( @xmath77 ) integration
( 6 ) , the operator @xmath93 will enter into truncation uncertainties @xmath100 by various combinations .
they can be extracted similarly as for @xmath59 , and we come to a force - gradient decomposition approach .
the most general representation of this approach is @xmath101 where again at a given @xmath78 the coefficients @xmath43 , @xmath44 , as well as @xmath102 must be chosen in such a way to cancel the truncation terms @xmath47 to the highest possible order @xmath2 . for @xmath103 , generalized factorization ( 13 )
reduces to usual representation ( 6 ) .
it is worth emphasizing that in view of the velocity independence of @xmath95 on @xmath50 , the modified operator of shifting velocities remains to be evaluated exactly for any @xmath44 and @xmath102 , namely , @xmath104 for quantum systems , where @xmath105 of solutions ( following from the property @xmath106 of evolution operator @xmath107 ) can be reproduced exactly as well by imposing additional constraints on the coefficients @xmath43 , @xmath44 , and @xmath102 . in particular , for velocity - like decompositions such constraints read : @xmath62 , @xmath108 , @xmath109 , and @xmath110
. then single - exponential subpropagators will enter symmetrically into the decompositions , providing automatically the required reversibility .
the case when the operators of shifting velocity and position are replaced by each other in the resulting symmetrical decomposition is also possible .
this leads to a position - like integration which can be reproduced from eq .
( 13 ) at @xmath111 , @xmath112 , and @xmath113 at @xmath114 and @xmath115 .
the above symmetry will result in its turn to automatic disappearing all even - order terms in the error function @xmath47 .
for this reason , the order @xmath2 of time - reversible ( self - adjoint ) algorithms may accept only even numbers ( @xmath116 ) .
the cancellation of the remaining odd - order terms up to a given order will be provided by fulfilling a set of basic conditions for @xmath43 , @xmath44 , and @xmath102 .
for example , the condition @xmath117 is required to cancel the first - order truncation uncertainties .
then the error function can be cast in the form @xmath118 \\ [ -5pt ] & & \ldots + { \cal o}_{k+1 } \delta t^{k+1 } \ , .
\nonumber\end{aligned}\ ] ] in order to kill higher odd - order truncation terms in eq .
( 15 ) , let us write down explicit expressions for @xmath119 , @xmath120 , and @xmath121 ( this will be enough to derive algorithms up to the eighth order ) .
expanding both the sides of eq .
( 13 ) into taylor s series , and collecting the terms with the same powers of @xmath25 one finds : @xmath122 + \beta [ { \cal b},[{\cal a},{\cal b } ] ] \ , , \ ] ] @xmath123 ] ] + \gamma_2 [ { \cal a},[{\cal a},[{\cal b},[{\cal a},{\cal b } ] ] ] ] + \gamma_3 [ { \cal b},[{\cal a},[{\cal a},[{\cal a},{\cal b } ] ] ] ] + \gamma_4 [ { \cal b},[{\cal b},[{\cal a},[{\cal a},{\cal b } ] ] ] ] \ , , \ ] ] @xmath124 ] ] ] ] + \zeta_2 [ { \cal b},[{\cal b},[{\cal b},[{\cal a},[{\cal a},[{\cal b},{\cal a } ] ] ] ] ] ] & + & \zeta_3 [ { \cal b},[{\cal b},[{\cal a},[{\cal a},[{\cal a},[{\cal b},{\cal a } ] ] ] ] ] ] + \nonumber \\ \zeta_4 [ { \cal b},[{\cal a},[{\cal b},[{\cal a},[{\cal a},[{\cal b},{\cal a } ] ] ] ] ] ] + \zeta_5 [ { \cal a},[{\cal b},[{\cal b},[{\cal a},[{\cal a},[{\cal b},{\cal a } ] ] ] ] ] ] & + & \zeta_6 [ { \cal a},[{\cal b},[{\cal a},[{\cal b},[{\cal a},[{\cal b},{\cal a } ] ] ] ] ] ] + \\ \zeta_7 [ { \cal b},[{\cal a},[{\cal a},[{\cal a},[{\cal a},[{\cal b},{\cal a } ] ] ] ] ] ] + \zeta_8 [ { \cal a},[{\cal b},[{\cal a},[{\cal a},[{\cal a},[{\cal b},{\cal a } ] ] ] ] ] ] & + & \zeta_9 [ { \cal a},[{\cal a},[{\cal b},[{\cal a},[{\cal a},[{\cal b},{\cal a } ] ] ] ] ] ] + \nonumber \\ \zeta_{10 } [ { \cal a},[{\cal a},[{\cal a},[{\cal a},[{\cal a},[{\cal b},{\cal a } ] ] ] ] ] ] \ , . \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \nonumber\end{aligned}\ ] ] 2 here we take into account the fact that the operators @xmath36 and @xmath93 commute between themselves , i.e. @xmath125=0 $ ] , so that any occurrence of constructions containing the chain @xmath126]$ ] has been ignored ( in particular for fifth - order truncation term @xmath120 this has allowed us to exclude the two zero - valued commutators @xmath127]]$ ] and @xmath128]]$ ] ) .
the multipliers @xmath129 , @xmath130 , @xmath131 , and @xmath132 , arising in eq .
( 16)(18 ) , are functions of the coefficients @xmath43 , @xmath44 , and @xmath102 , where @xmath133 .
the concrete form of these functions will depend on @xmath78 and the version ( velocity or position ) under consideration .
the most simple way to obtain explicit expressions for the multipliers consists in the following .
first , since we are dealing with self - adjoint schemes , the total number of single - exponential operators ( stages ) in eq .
( 13 ) is actually equal to @xmath134 , i.e. it accepts only odd values ( mention that one of the boundary set of coefficients is set to zero , @xmath62 or @xmath135 ) .
then we can always choose a central single - exponential operator , and further consecutively applying @xmath136 times the two types of symmetric transformation @xmath137
@xmath138 @xmath139 & & { \rm e}^{{\cal b } b^{(n ) } \delta t+{\cal c } c^{(n ) } \delta t^3 } { \rm e}^{{\cal w}^{(n ) } + { \cal o}^{(n ) } } { \rm e}^{{\cal b } b^{(n ) } \
\delta t+{\cal c } c^{(n ) } \delta t^3}\end{aligned}\ ] ] come to factorization ( 13 ) , where @xmath140 and @xmath141 is defined by eq .
the quantities @xmath142 , @xmath143 , and @xmath144 are related to @xmath43 , @xmath44 , and @xmath102 , respectively ( the relationship between @xmath145 and @xmath146 is determined below ) . for velocity - like decomposition with even @xmath78 or position - like at odd @xmath78 ,
the central operator is correspondingly @xmath147 or @xmath148 . so that here we must put @xmath149 as well as @xmath150 and let either @xmath151 or @xmath152 on the very beginning ( @xmath153 ) of the recursive procedure .
the start of the procedure should be performed with the second line of eq .
( 19 ) at @xmath154 and @xmath155 or @xmath156 and @xmath157 with further decreasing the index @xmath146 with increasing the number @xmath158 at @xmath159 , @xmath160 , and @xmath161 in both the lines of transformation ( 19 ) . for velocity - like decomposition with odd @xmath78 or position - like at even @xmath78 ,
the central operator will be @xmath162 or @xmath163 , corresponding to @xmath164 and @xmath165 or @xmath166 and @xmath167 , respectively , with @xmath168 and @xmath169 . in this case , the procedure should be started with the first type of transformation at @xmath170 or @xmath171 with decreasing @xmath146 at increasing @xmath145 for @xmath160 , @xmath172 , and @xmath159 in eq .
( 19 ) .
the recursive relations between the multipliers @xmath173 , @xmath174 , @xmath129 , @xmath130 , and @xmath131 corresponding to the first line of eq .
( 19 ) are : @xmath175 @xmath176 @xmath177 @xmath178 @xmath179 @xmath138 @xmath180 @xmath181 2 for the second type of transformation the relations read : @xmath182 @xmath183 @xmath184 @xmath185 @xmath186 @xmath187 @xmath188 & & 3 { b^{(n)}}^3 \nu^{(n ) } + 30 c^{(n ) } \sigma^{(n ) } - 2 { b^{(n)}}^2 \nu^{(n ) } \sigma^{(n ) } - b^{(n ) } \nu^{(n ) } { \sigma^{(n)}}^2\big)\big)/180 \ , .\end{aligned}\ ] ] 2 the relations for @xmath132 are presented in appendix . in such a way , at the end of the recursive process ( i.e. after @xmath136 steps ) the multipliers can readily be obtained .
the form of the first two multipliers are particularly simple and look as @xmath189 and @xmath190 .
so that , as was already mentioned above , putting @xmath191 and @xmath192 will cancel the first - order truncation uncertainties ( because the resulting exponential propagator must behave like @xmath193 ) .
next multipliers should be set to zero and we come to the necessity of solving a system of non - linear equations ( so - called order conditions ) with respect to @xmath43 , @xmath44 , and @xmath102
. we shall now consider actual self - adjoint algorithms of orders @xmath59 , 4 , 6 , and 8 .
putting @xmath61 in eq .
( 13 ) with @xmath62 , @xmath194 , @xmath64 , and @xmath195 leads to the following velocity - force - gradient algorithm of the second ( @xmath59 ) order , @xmath196 with @xmath197 , and @xmath198 .
note that here and below , for reducing the number of unknowns , we will always take into account in advance the symmetry of coefficients @xmath43 , @xmath44 , and @xmath102 as well as the fulfilling the first - order conditions @xmath199 when writing decomposition formulas . then solving the equation @xmath200 yields @xmath201 and we come to the already found integrator ( 12 ) .
it is worth remarking that negative values of quantities @xmath102 at force gradients have nothing to do with the above problem of positiveness of time coefficients arising at velocities and forces , i.e. , for @xmath43 and @xmath44 .
the reason is that the incremental velocity @xmath202 in eq .
( 14 ) can be rewritten as @xmath203 , and thus treated as the velocity changing in a modified step - size - dependent acceleration field @xmath204 .
the position counterpart of eq .
( 28 ) is obtained from eq . ( 13 ) at @xmath62 , @xmath205 , @xmath206 , @xmath207 , @xmath208 and @xmath209 , that yields @xmath210 for which @xmath211 and @xmath212 . letting @xmath213 will minimize the third - order truncation errors to the value @xmath214 \delta t^3 $ ] which is even twice smaller in magnitude than that of the velocity version .
note , however , that for both versions ( 28 ) and ( 29 ) , which require one force plus one force - gradient evaluations per time step , the order of integration is not increased with respect to the usual ( when @xmath215 ) verlet integrators requiring only one force recalculation . in view of this , the applying force gradients in a particular case of @xmath61 can be justified only for strongly interacting systems when the kinetic part @xmath35 of the liouville operator @xmath29 is much smaller than the potential part @xmath36 , i.e. , when @xmath216 with @xmath217 . then the remaining part @xmath218
\delta t^3 $ ] of local uncertainties will behave like @xmath219 and can be neglected .
further increasing @xmath78 on unity allows us to kill exactly both the multipliers @xmath129 and @xmath130 , that is needed for obtaining fourth - order ( @xmath65 ) integrators . so that choosing @xmath220 leads to the velocity - like propagation @xmath221 \\ [ 2pt ] & & { \rm e}^{{\cal b } \lambda \delta t + \xi { \cal c } \delta t^3 } { \rm e}^{{\cal a } \frac{\delta t}{2 } } { \rm e}^{{\cal b } ( 1 - 2 \lambda ) \delta t + \chi { \cal c } \delta t^3 } { \rm e}^{{\cal a } \frac{\delta t}{2 } } { \rm e}^{{\cal b } \lambda \delta t + \xi { \cal c } \delta t^3 } \ \ \ \ \nonumber\end{aligned}\ ] ] following from eq .
( 13 ) at @xmath62 , @xmath222 , @xmath223 , @xmath224 , @xmath225 , and @xmath226 . here relations ( 21 ) , ( 22 ) , ( 25 ) , and ( 26 ) come to the two order conditions @xmath227 with three unknowns @xmath228 , @xmath229 , and @xmath230 .
the first unknown is immediately obtained satisfying the first condition , @xmath231 the second equality is then reduces to @xmath232 , resulting in a whole family of velocity - force - gradient algorithms of the fourth order . in general , such algorithms will require two force and two force - gradient recalculations per time step . remembering that we are interested in the derivation of most efficient integrators , three cases deserve to be considered .
the two of them are aimed to reduce the number of force - gradient recalculations from two to one .
this is possible by choosing either @xmath233 or @xmath234 in the third case we will try to minimize the norm @xmath235 of fifth - order truncation errors @xmath236 at @xmath237 and @xmath238 , treating @xmath229 as a free parameter . in view of recursive relations ( 23 ) and
( 27 ) , explicit expressions for the components of @xmath236 are @xmath239 @xmath240 then taking into account eq .
( 31 ) one finds the function @xmath241 with the minimum @xmath242 at @xmath243 at the same time , the values of @xmath244 corresponding to first two algorithms ( 32 ) and ( 33 ) constitute @xmath245 and @xmath246 , respectively . position version of ( 30 ) reads @xmath247 \\ [ -5pt ] & & { \rm e}^{{\cal a } \lambda \delta t } { \rm e}^{{\cal b } \frac{\delta t}{2 } + \xi { \cal c } \delta t^3 } { \rm e}^{{\cal a } ( 1 - 2 \lambda ) \delta t } { \rm e}^{{\cal b } \frac{\delta t}{2 } + \xi { \cal c } \delta t^3 } { \rm e}^{{\cal a } \lambda \delta t } \nonumber \ \ \ \\end{aligned}\ ] ] and is obtained from eq .
( 13 ) at @xmath220 with @xmath248 , @xmath249 , @xmath194 , @xmath250 , and @xmath251 . here
, the number of unknowns coincides with the number of the order conditions @xmath252 solving of which yields two solutions , @xmath253 then the norm of truncation uncertainties @xmath236 appearing in eq .
( 36 ) is @xmath254 , so that the preference should be given to sign `` '' in eq .
( 37 ) , because this leads to a smaller value , @xmath255 , of @xmath244 ( whereas @xmath256 ) .
position algorithm ( 36 ) needs , like velocity version ( 35 ) , in two force and the same number of force - gradient evaluations per time step . integrators ( 32 ) and ( 37 ) have been previously derived by suzuki @xcite based on mclachlan s method of small perturbation @xcite and referred by chin @xcite to schemes a and b , respectively . algorithms ( 33 ) and ( 35 ) are new and will labeled by us as a@xmath257 and a@xmath258 . while scheme a@xmath257 seems has no advantages over the a - integrator , the new algorithm a@xmath258 corresponds to the best accuracy of the integration , because it minimizes @xmath244 .
it requires , however , one extra force - gradient evaluation and , thus , can be recommended for situations when this evaluation does not present significant difficulties . with the aim of considerable decreasing the truncation errors in a little additional computational efforts , chin @xcite
has proposed to consider extended force - gradient algorithms of the fourth order .
this has been achieved by increasing the number of force recalculations on unity with respect to the necessary minimum , i.e. choosing @xmath259 . at the same time , the number of force - gradient evaluations was fixed to its minimal value @xmath260 . within our general approach ,
it is possible to introduce two fourth - order schemes satisfying the above requirements .
the schemes are @xmath261 \\ [ -6pt ] \times { \rm e}^{{\cal b } ( 1 - 2 \lambda ) \delta t + \chi
{ \cal c } \delta t^3 } { \rm e}^{{\cal a } ( 1 - 2 \theta ) \frac{\delta t}{2 } } { \rm e}^{{\cal b } \lambda \delta t } { \rm e}^{{\cal a } \theta \delta t } & & \nonumber\end{aligned}\ ] ] and @xmath262 \\ [ -6pt ] \times { \rm e}^{{\cal a } ( 1 - 2 \theta ) \delta t } { \rm e}^{{\cal b } ( 1 - 2 \lambda ) \frac{\delta t}{2 } } { \rm e}^{{\cal a } \theta \delta t } { \rm e}^{{\cal b } \lambda \delta t + \xi { \cal c } \delta t^3 } \nonumber\end{aligned}\ ] ] following from eq .
( 13 ) at @xmath66 and corresponding to position- and velocity - like integration , respectively .
note that further we will not present the relationship between the coefficients @xmath43 , @xmath44 , @xmath102 of eq .
( 13 ) and reduced variables ( such as , for example , @xmath263 , @xmath228 , @xmath230 in eq . ( 38 ) ) in view of its evidence .
the order conditions for scheme ( 38 ) are @xmath264 \\ [ -2pt ] \beta&=&-\frac{1}{12 } + \frac{\lambda}{2 } - \frac{\lambda^2}{2 } - \lambda \theta \big ( 1 - \lambda \big ) + \chi = 0\end{aligned}\ ] ] and solving them one obtains @xmath265 relations ( 40 ) constitute a family of extended force - gradient position algorithms ( 38 ) of the fourth order with @xmath228 being a free parameter .
chin @xcite has introduced an algorithm like ( 38 ) in somewhat another way , namely , as a symmetric product of two third - order schemes .
this results only in one set of time coefficients which can be reproduced ( at sign `` '' ) from eq .
( 40 ) as a particular case corresponding to @xmath266 and has been referred to scheme c. solution ( 41 ) may not , however , be necessarily optimal in view of the fact that it does not minimize the norm @xmath244 ( see eq .
( 34 ) ) of truncation uncertainties @xmath236 . indeed ,
the components of @xmath244 for scheme ( 38 ) are @xmath267 \gamma_3&=&-\frac{1}{360 } \bigg ( \frac{3}{2 } \pm \frac{5}{\sqrt{96 \lambda } } \pm 5 \sqrt{\frac{\lambda}{24 } } \bigg ) \
, , \\ [ 2pt ] \gamma_4&=&-\frac{1}{1440 } \big ( 3 \pm 5 \sqrt{24 \lambda } + 45 \lambda - 30 \lambda^2 \big ) \ , , \end{aligned}\ ] ] where eq
( 40 ) has been used to express the function @xmath268 in terms of one parameter @xmath228 exclusively .
the global minimum of this function is @xmath269 and achieved ( at sign minus ) at @xmath270 ( all results found numerically will be presented within sixteen significant digits for schemes up to the eighth order and within thirty two digits for order ten and higher ) . on the other hand , the value of @xmath244 corresponding to scheme c ( eq . ( 41 ) ) ,
is equal only to @xmath271 , i.e. it is approximately in 5 times larger than that of the optimized algorithm ( 42 ) .
the last algorithm we will designate as scheme c@xmath257 . a similar pattern is observed in the case of extended velocity - force - gradient integration ( 39 ) .
previously chin and chen @xcite have indicated that for quantum mechanics simulations the integration of such a type is more preferable than position - like scheme ( 38 ) , because it requires a fewer number of spatial fourier transforms . again using the symmetric product of two third - order integrators to increase the order from three to four
, they have obtained the following set @xmath272 of time coefficients and referred it to scheme d. we have realized that this set is not only possible and found a whole family of solutions ( which includes ( 43 ) ) , namely , @xmath273 where @xmath263 should be considered as a free parameter .
the optimal solution , which minimizes the norm @xmath244 of fifth - order errors to the value @xmath274 , is @xmath275 and will be labeled as scheme d@xmath257 . at the same time , the norm of errors corresponding to scheme d ( eq . ( 43 ) ) is equal to @xmath276 , i.e. it exceeds the minimum , that may results in decreasing the precision of the calculations .
as can be ensured readily , the time coefficients arising at basic operators @xmath35 and @xmath36 under exponentials are positive for all the fourth - order force - gradient algorithms described in this subsection . therefore , contrary to usual force forest - ruth - like schemes , such algorithms can simulate dynamical processes in all areas of physics and chemistry without any principal restrictions .
beginning from @xmath277 , the force - gradient factorization being written in velocity representation allows to eliminate the components of truncation uncertainties up to the sixth order ( @xmath72 ) inclusively . in view of eq .
( 13 ) , such a representation reads @xmath278 the number of unknowns in propagation ( 45 ) is the same as the number of order conditions which now take the form @xmath279 @xmath280 @xmath281 @xmath282 \\ [ -8pt ] & & 20 \lambda ( 2 \theta -1 ) ( \theta + 2 \vartheta - 6 \theta \vartheta ) + 80 \xi - 480 \theta \xi + 480 \theta^2 \xi \bigg ) = 0 \ , , \nonumber\end{aligned}\ ] ] @xmath283 @xmath284 2 the unique real solution to system ( 46 ) is @xmath285{675 + 75 \sqrt{6}}}{30 } + \frac{5}{2 \sqrt[3]{675 + 75 \sqrt{6 } } } \ , , \ \ \
\vartheta = \frac{\theta}{3 } \ , , \ ] ] @xmath286 @xmath287 solution ( 47 ) constitutes a velocity - force - gradient algorithm of the sixth order with four force and three ( since @xmath288 ) force - gradient evaluations per time step , i.e. , with @xmath289 and @xmath290 .
its advantage over usual sixth - order integrators consists in the fact that it is composed from a considerably smaller number , namely @xmath291 , instead of 15 , of single exponential operators .
the norm @xmath292 of seventh - order truncation errors @xmath293 ( see eq .
( 45 ) ) , corresponding to solution ( 47 ) , is equal to @xmath294 .
note also that the position version of decomposition ( 45 ) does not exist at @xmath277 , because then the number of unknowns is less than the number of order equations , resulting in the absence of solutions . as has been shown in the preceding subsection for the case of fourth - order integration , algorithms with minimal numbers @xmath295 of force evaluation may not lead to optimal solutions .
the reason is that slight increasing @xmath295 may significantly decrease the local errors and thus overcompensate an increased computational efforts . so that increasing @xmath295 as well as @xmath78 on unity ( note that @xmath296 ) and do not changing @xmath297 , i.e. choosing @xmath298 with @xmath299 and @xmath300 , it is possible to derive from decomposition ( 13 ) up four ( two velocity- and two position - like ) extended sixth - order schemes .
they are @xmath301 \\ [ -7pt ] { \rm abacacacaba } \ , , \ \ \ \ \ \ { \rm acabacabaca } \ , , \nonumber\end{aligned}\ ] ] where we have used an abbreviation that a and b denote exponential operators @xmath302 and @xmath303 , respectively , whereas letter c corresponds to @xmath304 .
each of these extended schemes has itself correspondingly six , eight , four , and two sets of real solutions for time coefficients .
we have realized that the smallest values of the norm @xmath305 ( see eq . ( 48 ) ) of local errors @xmath293 within the sets are 0.0000264 , 0.0000147 , 0.000146 , and 0.00000607 , respectively . so that the last scheme should be considered as the best .
more explicit form for it is @xmath306 with the optimal solution @xmath307 \\ [ -7pt ] \lambda = \ \ 0&.&1131980348651556{\rm
e}\!+\!01 \nonumber \\ \chi = \ ! -0&.&1324638643416052{\rm
e}\!-\!01 \nonumber \\ \mu = \ \ 0&.&8642161339706166{\rm e}\!-\!03 \nonumber\end{aligned}\ ] ] corresponding to @xmath308 .
in such a way , the error function has been reduced more than in 200 times with respect to scheme ( 47 ) for which @xmath294 . in the case when @xmath74 we must satisfy up eighteen order conditions , namely , @xmath191 , @xmath192 , @xmath309 , @xmath200 , @xmath310 , and @xmath311 .
taking into account the symmetry of time coefficients @xmath43 , @xmath44 , and @xmath102 , this can be achieved at least at @xmath312 , i.e. , using @xmath313 single exponential operators .
for @xmath312 the velocity- and position - like force - gradient decomposition ( 13 ) transforms into the schemes @xmath314 respectively . the number of unknowns for both the schemes are also eighteen and we can try to solve the system of order conditions with respect to these unknowns .
it is worth remarking such a system appears to be very cumbersome for schemes under consideration .
for instance , the resulting non - linear equations of this system being written explicitly in mathematica create a file of 0.5 mb in length ! in view of this , our attempts to solve the equations symbolically have not meet with much success .
we mention that all the results presented above for algorithms of orders 2 , 4 , and 6 have been solved analytically or in quadratures . saying in quadratures we mean that the problem was reduced to finding real zeros for a one - dimensional polynomial of a given order . so that we could identify exactly the number of solutions and their locations . here
the situation is somewhat different because we must solve the system using purely numerical approaches , such as the newton method . as a result
, one can not guarantee that we will found all possible solutions .
however , solving the system on a computer during significantly long time , one can stay with a great probability that we have found almost all physically interesting solutions and chosen among them nearly optimal sets .
the numerical calculations has been performed in fortran using the well - recognized newton solver with numerical determination of partial derivatives .
the values for non - linear functions ( that constitute the system of equations ) were obtained using recursive relations ( 20)(27 ) , ( a1 ) , and ( a2 ) , but not explicit expressions for them to save the processor time and increase the precision of the computations .
the initial guess for solutions were generated at random within the interval @xmath315 $ ] in each the eighteenth directions . if newton s iterations become to diverge at a particular guess or during the calculations , a next random point was involved to repeat the process .
in such a way , after several days of continuous attacking the systems of equations on an origin 3800 workstation , we found two and five solutions for schemes ( 52 ) and ( 53 ) , respectively .
the optimal among them are following @xmath316 b_1 = b_{12 } = \ \ 0&.&1839699354244402{\rm e}\!+\!00 \nonumber \\ [ -2pt ] c_1 = c_{12 } = \ \ 0 & & \nonumber \\ [ -2pt ] a_2 = a_{12 } = \ \ 0&.&6922517172738832{\rm e}\!+\!00 \nonumber \\ [ -2pt ]
b_2 = b_{11 } = \ \ 0&.&7084389757230299{\rm e}\!+\!00 \nonumber \\ [ -2pt ] c_2 = c_{11 } = \ \ 0&.&3976209968238716{\rm e}\!-\!01 \nonumber \\ [ -2pt ] a_3 = a_{11 } = -0&.&3183450347119991{\rm
e}\!+\!00 \nonumber \\ [ -2pt ] b_3 = b_{10 } = \ \ 0&.&1981440445033534{\rm e}\!+\!00 \nonumber \\ [ -2pt ] c_3 = c_{10 } = \ \ 0&.&2245403440322733{\rm e}\!-\!01 \nonumber \\ [ -2pt ] a_4 = a_{10 } = \ \ 0&.&6766724088765565{\rm e}\!+\!00 \nonumber \\ [ -2pt ] b_4 = b_9 = -0&.&6409380745116974{\rm e}\!-\!01 \nonumber \\ [ -2pt ] c_4 = c_9 = \ \
0&.&9405266232181224{\rm e}\!-\!03 \nonumber \\ [ -2pt ] a_5 = a_9 = -0&.&7207972470858706{\rm e}\!+\!00 \nonumber \\ [ -2pt ] b_5 = b_8 = -0&.&6887429532761409{\rm e}\!+\!00 \nonumber \\ [ -2pt ] c_5 = c_8 = -0&.&7336500519635302{\rm e}\!-\!01 \nonumber \\ [ -2pt ] a_6 = a_8 = \ \
0&.&3580316862350045{\rm e}\!+\!00 \nonumber \\ [ -2pt ] b_6 = b_7 = \ \
0&.&1622838050764871{\rm e}\!+\!00 \nonumber \\ [ -2pt ] c_6 = c_7 = \ \
0&.&2225664796363730{\rm e}\!-\!01 \nonumber \\ [ -2pt ] a_7 = -0&.&3756270611751488{\rm e}\!+\!00 \nonumber\end{aligned}\ ] ] for velocity - like integration ( 52 ) , and @xmath317 c_{12 } = \ \ 0 & & \nonumber \\ [ -2pt ] a_1 = a_{12 } = \ \ 0&.&41009674738801111928784693005080{\rm e}\!+\!00 \nonumber \\ [ -2pt ] b_1 = b_{11 } = \ \ 0&.&48249309817414952912695842664785{\rm e}\!-\!02 \nonumber \\ [ -2pt ] c_1 = c_{11 } = \ \ 0&.&14743936907797528364717244760736{\rm e}\!-\!03 \nonumber \\ [ -2pt ] a_2 = a_{11 } = -0&.&34123345756052780489101697378499{\rm
e}\!+\!00 \nonumber \\ [ -2pt ] b_2 = b_{10 } = \ \ 0&.&17492394861090375603419001374207{\rm e}\!+\!00 \nonumber \\ [ -2pt ] c_2 = c_{10 } = \ \ 0&.&23288450531932545357194967600155{\rm e}\!-\!03 \nonumber \\ [ -2pt ] a_3 = a_{10 } = \ \
0&.&25644714021068150492361761631743{\rm e}\!+\!00 \nonumber \\ [ -2pt ] b_3 = b_9 = \ \
0&.&29304366370957066164364546204288{\rm e}\!+\!00 \nonumber \\ [ -2pt ] c_3 = c_9 = \ \ 0&.&61648659635535962497705619884752{\rm e}\!-\!02 \nonumber \\ [ -2pt ]
a_4 = a_9 = \ \
0&.&27765273975812438394100476242641{\rm e}\!+\!00 \nonumber \\ [ -2pt ] b_4 = b_8 = \ \
0&.&47448940168459770284238136482511{\rm e}\!-\!01 \nonumber \\ [ -2pt ] c_4 = c_8 = -0&.&12307516860831240716732016960034{\rm e}\!-\!01 \nonumber \\ [ -2pt ] a_5 = a_8 = -0&.&56926266869753773902939657321159{\rm e}\!+\!00 \nonumber \\ [ -2pt ] b_5 = b_7 = -0&.&15299863411743974499219652320477{\rm e}\!-\!02 \nonumber \\ [ -2pt ] c_5 = c_7 = -0&.&73296648559126385387017161643798{\rm e}\!-\!04 \nonumber \\ [ -2pt ] a_6 = a_7 = \ \ 0&.&46629949890124853576794423820194{\rm e}\!+\!00 \nonumber \\ [ -2pt ]
b_6 = -0&.&37422994259002571606842462603791{\rm e}\!-\!01 \nonumber \\ [ -2pt ] c_6 = \ \ 0&.&15295860994523744731993293847001{\rm e}\!-\!01 \nonumber\end{aligned}\ ] ] for its position - like counterpart ( 53 ) .
the number of force evaluations per times step for schemes ( 52 ) and ( 53 ) is @xmath318 , whereas the number of force - gradient recalculations consists @xmath319 ( since @xmath320 and thus the two boundary letters c in formula ( 52 ) should be actually replaced by b ) and @xmath321 , respectively . in view of a complicated structure of the ninth - order truncation uncertainties @xmath322 , the optimal solutions just presented
have been chosen in somewhat other way than above , namely , by providing a minimum for the function @xmath323 .
this simplified criterion was used , in particular , by kahan and li @xcite , when optimizing usual force algorithms . as a result , we have obtained @xmath324 for scheme ( 52 ) and @xmath325 is smaller in the last case , the position - like integration should be considered as more preferable .
its time coefficients have been presented even with thirty second significant digits to be used in applications for very accurate integration . in order to ensure that all the digits shown are correct in both the cases , we have carried out a few additional newton s iterations in maple with up 200 digits during the internal computations , and taking as initial guesses the solutions already obtained in fortran .
the position - like decomposition ( 53 ) has also another advantage over the velocity version ( 52 ) in that the all the intermediate ( @xmath326 ) states in velocity and position space stay during the integration within a given interval @xmath327 $ ] , i.e. , @xmath328 and @xmath329 .
this property may be important when solving ordinary differential equations ( for specific physical systems or in pure mathematics applications ) with singularities beyond the interval of the integration ( note that , in particular , any system of differential equations of the form @xmath330 is reduced to the equations of motion under consideration in this paper ) .
note also that in order to construct eighth - order schemes within usual decomposition approach ( 6 ) ( i.e. , without involving any force - gradients ) , it could be necessary to apply up @xmath331 ( instead of 23 ) single exponential propagators .
such schemes has never been derived by decomposition ( 6 ) because of the serious technical difficulties .
they can be explicitly introduced only by compositions of lower - order integrators ( see the next section ) . instead ,
using generalized scheme ( 13 ) has allowed us to derive eighth - order algorithms by direct decompositions for the first time ( the force - gradient algorithms presented in subsection ii.b._5 _ for order six are completely new as well ) .
all the decomposition algorithms obtained by us in subsections ii.b._3 _ , _ 4 _ , _ 5 _ , and _ 6 _ are collected below in table 1 . here
, the designations err3 , err5 , and err7 relate to the norms @xmath332 , @xmath244 , and @xmath305 of correspondingly third- , fifth- , and seventh - order truncation errors ( see eqs . ( 6 ) , ( 15)(18 ) , ( 34 ) , and ( 48 ) ) , whereas @xmath295 and @xmath295 denote the numbers of force and force - gradient evaluations per time step .
the optimal algorithms for orders 2 , 4 , 6 , and 8 are labeled by g2 , c@xmath257 , g6 , and g8 , respectively . among other schemes presented for each given order ,
such algorithms reduce the truncation uncertainties to a minimum . taking into account that this reduction is achieved at the same or nearly the same computational efforts ,
the optimal algorithms should be considered as the best not only with respect to their precision but in view of the overall efficiency as well ( see also comments on this in section iii ) .
lccccccccl algorithm & order & @xmath295 & @xmath297 & err3 & err5 & err7 & equations & remarks & @xmath333 label + + [ -8pt ] cac & 2 & 1 & 1 & @xmath334 & @xmath335 & @xmath336 & ( 28 ) & new & g2@xmath257 + aca & 2 & 1 & 1 & @xmath337 & @xmath338 & @xmath339 & ( 29 ) & new@xmath340 & g2 + + [ -8pt ] bacab & 4 & 2 & 1 & 0 & @xmath341 & @xmath342 & ( 30,32 ) & refs .
@xcite & a + cabac & 4 & 2 & 1 & 0 & @xmath343 & @xmath344 & ( 30,33 ) & new & a@xmath257 + cacac & 4 & 2 & 2 & 0 & @xmath345 & @xmath346 & ( 30,35 ) & new & a@xmath258 + acaca & 4 & 2 & 2 & 0 & @xmath347 & @xmath348 & ( 36,37 ) & refs .
@xcite & b + abacaba & 4 & 3 & 1 & 0 & @xmath349 & @xmath350 & ( 38,41/42 ) & ref .
@xcite / new@xmath340 & @xmath351 c / c@xmath257 + cababac & 4 & 3 & 1 & 0 & @xmath352 & @xmath353 & ( 39,43/44 ) & ref .
@xcite / new & @xmath351 d / d@xmath257 + + [ -8pt ] bacacacab & 6 & 4 & 3 & 0 & 0 & @xmath354 & ( 45,47 ) & new & g6@xmath257 + cacababacac & 6 & 5 & 3 & 0 & 0 & @xmath355 & ( 49 ) & new & g6@xmath258 + cabacacabac & 6 & 5 & 3 & 0 & 0 & @xmath356 & ( 49 ) & new & g6@xmath357 + abacacacaba & 6 & 5 & 3 & 0 & 0 & @xmath358 & ( 49 ) & new & g6@xmath359 + acabacabaca & 6 & 5 & 3 & 0 & 0 & @xmath360 & ( 50,51 ) & new@xmath340 & g6 + + [ -8pt ] bacacacacaca + [ -6.5pt ] & 8 & 11 & 10 & 0 & 0 & 0 & ( 52 ) & new & g8@xmath257 + [ -6.5pt ] @xmath361cacacacacab + [ 2.5pt ] acacacacacac + [ -6.5pt ] & 8 & 11 & 11 & 0 & 0 & 0 & ( 53 ) & new@xmath340 & g8 + [ -6.5pt ] @xmath361acacacacaca @xmath362 the best algorithm within a given order @xmath363the value corresponding to scheme c@xmath257
@xmath364the value corresponding to scheme d@xmath257 2 finally , it is worth remarking that the problem of constructing algorithms with only positive coefficients @xmath43 and @xmath44 for orders six and higher still remains .
we mention that for order four , this problem has been resolved ( see subsection ii.b._4 _ ) by transferring the force - gradient component of truncation uncertainties into the exponential propagators . for orders @xmath365
, additional higher - order gradients should appear under these exponentials to provide the required positiveness .
our analysis has shown , however , that such high - order exponentials ( besides their very cumbersome forms ) can not be evaluated in quadratures and need in performing implicit calculations by iteration . in view of this
we can come to a conclusion that beyond fourth order , analytically integrable decomposition algorithms with purely positive coefficients do not exist .
mathematically rigorous prove of this statement will be considered in our further investigation and presented elsewhere . with increasing the order of integration to ten and higher
, the construction of algorithms by direct decompositions ( 13 ) becomes to be inefficient because of a large number of the order conditions and time coefficients .
however , having the already derived force - gradient integrators of lower orders @xmath2 , we can try to compose them as @xmath366 for obtaining an algorithm of order @xmath367 . then the composition constants @xmath368 , where @xmath133 , should be chosen in such a way to provide the maximal possible value of @xmath369 at a given number @xmath370 .
note that lower - order propagations @xmath371 enter symmetrically in composition ( 54 ) and their total number @xmath372 accepts odd values . so that if a basic integrator @xmath373 is self - adjoint , the resulting algorithm @xmath374 will be self - adjoint as well .
the idea of using formula like ( 54 ) is not new and has been applied by different authors in previous investigations .
but these investigations were focused , in fact , on the compositions of usual second - order ( @xmath59 ) schemes ( to our knowledge , no actual calculations of composition constants for fourth- and higher - order - based integrators have been reported ) . although using the second - order - based approach allowed to introduce algorithms to the tenth order @xcite , further increasing @xmath369 has led to unresolved numerical difficulties when finding the coefficients of the compositions .
usually , these difficulties are obviated with the help of creutz s and gocksch s method @xcite .
we mention that according to this method , an algorithm of order @xmath1 can be derived by the triplet concatenation @xmath375 of a self - adjoint integrator of order @xmath2 , where @xmath376 .
in particular , chin and kidwell @xcite starting from force - gradient algorithm ( 41 ) of order four and repeating procedure ( 55 ) up to order 12 , have indicated a visible increasing the efficiency of the computation with respect to second - order - based schemes . in this approach
, however , the number of force and force - gradient evaluations ( the most time - consuming part of the calculations ) increases too rapidly with increasing @xmath2 , namely as @xmath377 relatively to the fourth - order integrator .
the present study is aimed to overcome the above problems by an explicit consideration of four- , sixth- , and eighth - order - based ( force - gradient ) algorithms within general composition approach ( 54 ) .
this results in reducing the total number of basic propagations to a minimum and providing significant speeding up the integration .
the composition algorithms are derived up to the sixteenth order inclusively . in the case when @xmath65 , the basic self - adjoint propagation is @xmath378 where @xmath379 .
explicit form of higher - order truncation operators @xmath380 , @xmath381 , @xmath382 , @xmath383 , and so on ( which was previously found for @xmath380 and @xmath381 , see eqs .
( 17 ) and ( 18 ) ) are not important within the composition approach . then formula ( 54 )
reduces to series ( @xmath384 ) of the transformation @xmath385 with @xmath386 and @xmath387 . in view of eqs .
( 56 ) and ( 57 ) , the structure of resulting propagation can be cast at each @xmath145 as @xmath388 with @xmath389 , \ ] ] @xmath390 \!+\ ! q_7 [ { \cal x}_1,{\cal x}_1,{\cal x}_1,{\cal x}_1,{\cal x}_5 ] , \ ] ] @xmath391 { \cal y}_{11 } = & & q_8 { \cal x}_{11 } + q_9 [ { \cal x}_1,{\cal x}_1,{\cal x}_9 ] + q_{10 } [ { \cal x}_1,{\cal x}_1,{\cal x}_1,{\cal x}_1,{\cal x}_7 ] + \\ & & q_{11 } [ { \cal x}_1,{\cal x}_1,{\cal x}_1,{\cal x}_1 , { \cal x}_1,{\cal x}_1,{\cal x}_5 ] + q_{12 } [ { \cal x}_5,{\cal x}_1,{\cal x}_5 ] .\end{aligned}\ ] ] comparing ( 56 ) and ( 58 ) yields values of @xmath392-multipliers at @xmath153 , namely , @xmath393 , @xmath394 , @xmath395 , @xmath396 , and @xmath397 , whereas , @xmath398 .
expanding both the sides of eq .
( 57 ) into taylor s series with respect to @xmath25 , one finds that values for these multipliers at @xmath399 can be obtained using the following recursive relations @xmath400 @xmath401 @xmath402 @xmath403 & & 7 q_2^{(n ) } { d^{(n)}}^2- { q_1^{(n)}}^3 { d^{(n)}}^k- 7 { q_1^{(n)}}^2 { d^{(n)}}^{k+1}-7 q_1^{(n ) } { d^{(n)}}^{k+2}\big)/360 \ , , \end{aligned}\ ] ] @xmath404 @xmath138 @xmath405 & & 7 q_3^{(n ) } { d^{(n)}}^2-{q_1^{(n)}}^3 { d^{(n)}}^{k+2}- 7 { q_1^{(n)}}^2 { d^{(n)}}^{k+3 } - 7 q_1^{(n ) } { d^{(n)}}^{k+4}\big)/360 \ , , \end{aligned}\ ] ] @xmath406 & & 42 { q_1^{(n)}}^2 q_2^{(n ) } { d^{(n)}}^2 + 294 q_4^{(n ) } { d^{(n)}}^2 - 62 q_1^{(n ) } q_2^{(n ) } { d^{(n)}}^3 + { q_1^{(n)}}^5 { d^{(n)}}^k - 31 q_2^{(n ) } { d^{(n)}}^4 + \\ [ 6pt ] & & 11 { q_1^{(n)}}^4 { d^{(n)}}^{k+1 } + 42 { q_1^{(n)}}^3 { d^{(n)}}^{k+2 } + 62 { q_1^{(n)}}^2 { d^{(n)}}^{k+3 } + 31 q_1^{(n ) } { d^{(n)}}^{k+4}\big)/15120 \ , , \end{aligned}\ ] ] @xmath407 2 applying the above relations @xmath136 times will give the final values of @xmath392-multipliers and thus lead to the desired order conditions .
for instance , the first condition is very simple and reads : @xmath408 .
this provides @xmath409 and guarantees ( see eqs . ( 56 ) , ( 58 ) and ( 59 ) ) that the order of the composition scheme will be at least not lower than that of the basic scheme , i.e. @xmath410 in our case .
all other multipliers @xmath411 , @xmath412 , @xmath413 ,
@xmath414 should be consecutively set to zero , forming higher - order conditions .
the total number @xmath415 of the conditions depends on a required order @xmath416 of the composition scheme .
in particular , at @xmath417 we must kill the term @xmath418 at fifth - order truncation uncertainties ( see eq . ( 58 ) ) .
taking into account eq .
( 59 ) , this results in two order conditions , namely , @xmath419 and @xmath420 which can be satisfied at @xmath61 .
then one obtains a system of equations , @xmath421 , and @xmath422 , with respect to two unknowns , @xmath423 and @xmath424 .
the system can be solved analytically , and the solution is @xmath425 with @xmath426 that coincides ( at @xmath65 ) with the result of triplet construction ( 55 ) .
this coinciding is not surprising because , as can be seen easily , both approaches ( 54 ) and ( 55 ) are identical in a partial case when @xmath61 and @xmath427 . with further increasing @xmath369 , composition approach ( 54 )
will lead to a more efficient integration .
indeed , choosing @xmath428 requires the term @xmath429 in eq . ( 58 ) should be killed additionally .
this is achieved by putting @xmath430 in eq .
( 59 ) , and , therefore , by solving at @xmath66 a system of four non - linear equations , @xmath419 , @xmath420 , @xmath431 , and @xmath432 with respect to the same number of unknowns @xmath423 , @xmath424 , @xmath433 , and @xmath434 . so that the minimal number of fourth - order integrators in the eight - order composition should be @xmath435 , whereas this number is equal to @xmath436 when triplet concatenation ( 55 ) is used .
expressions for the non - linear equations can readily be reproduced by applying the corresponding set of recursive relations ( 60 ) .
we will not present such expressions explicitly , because as has been realized , the order equations do not allow to be solved analytically at @xmath437 for any @xmath438 .
but , these equations can be solved in a quite efficient way numerically using the newton s method .
details of the numerical calculations are similar to those described in subsection ii.b._6_. here ( when @xmath66 , @xmath65 , and @xmath428 ) we have found five solutions , and it seems no other real solutions exist .
the optimal set is @xmath439 \\ [ -7pt ] d_3 = -0&.&1090206660543938{\rm e}\!+\!01 \nonumber \\
d_4 = \ \ 0&.&1172145334546811{\rm e}\!+\!01 \ , .
\nonumber\end{aligned}\ ] ] solution ( 61 ) simultaneously leads the smallest values for the maximal composition coefficient @xmath440 and the norm @xmath441 of the main ninth - order term @xmath442 of truncations uncertainties . when deriving tenth - order composition algorithms ( at @xmath65 ) , i.e. when @xmath443 , three additional order conditions arise , @xmath444 , @xmath445 , and @xmath446 , needed to eliminate the term @xmath442 ( see eqs .
( 58 ) and ( 59 ) ) .
then we come in overall to 7 non - linear equations which can be satisfied by appropriate choosing composition constants @xmath368 ( @xmath133 ) at @xmath447 . in this case
, we have identified more than 150 real solutions and probably there are somebody others ( we stopped the searching after several days of the computations ) . among the solutions found the optimal set looks @xmath448 d_2 = -0&.&49193105914623101022388138864143 \nonumber \\ [ -2pt ] d_3 = \ \
0&.&35449258654398460535529269988483 \nonumber \\ [ -8pt ] \\ [ -8pt ] d_4 = -0&.&69573922271140223803036463461997 \nonumber \\ [ -2pt ] d_5 = \
\ 0&.&39959538030329256359349977087819 \nonumber \\ [ -2pt ] d_6 = \ \ 0&.&54979568601438452794128031563760 \nonumber\end{aligned}\ ] ] and @xmath449 .
this set minimizes at once the norm @xmath450 of the main eleventh - order term @xmath451 of truncation errors and the quantity @xmath452 to the values 0.00412 and 0.843 ( @xmath453 ) , respectively . here
, the number of basic propagations ( stages ) is @xmath454 , i.e. more than in two times smaller than this number @xmath455 within triplet concatenation ( 55 ) . in order to introduce twelfth - order algorithms , @xmath456 , on the basis of fourth - order compositions it is necessary to deal with @xmath312 unknowns @xmath368 to fulfill the same number of the order conditions @xmath419 , and @xmath457=0 .
here we have found more than 200 real solutions and perhaps there are somebody else .
the best among them , which minimizes @xmath458 to the value 0.611 ( @xmath459 ) , is @xmath460 d_2 = \ \
0&.&53377479890712207949282653990842 \nonumber \\ [ -2pt ] d_3 = \
\ 0&.&12130138614668307673802291966495 \nonumber \\ [ -2pt ] d_4 = \
\ 0&.&29650747033807195273440032505629 \nonumber \\ [ -2pt ] d_5 = -0&.&59965999857335454018482312008233 \nonumber \\ [ -8pt ] \\ [ -8pt ] d_6 = \ \ 0&.&09043581286204437145871130429094 \ \nonumber \\ [ -2pt ] d_7 = -0&.&43979146257635806886778748138962 \nonumber \\ [ -2pt ] d_8 = -0&.&30251552922346495057010240779104 \nonumber \\ [ -2pt ] d_9 = \
\ 0&.&59895872989247982114545906953712 \nonumber \\ [ -2pt ] d_{10 } = \ \ 0&.&31236416538275576151816280776696 \nonumber \\ [ -2pt ] d_{11 } = -0&.&59081230769647833184090443445303 \nonumber\end{aligned}\ ] ] with @xmath461 .
thus , the minimal number of fourth - order stages needed to compose the twelfth - order algorithm is @xmath462 , instead of up @xmath463 as in the case of usual triplet construction ( 55 ) . when @xmath72 or 8 , the basic propagation reads @xmath464 or @xmath465 respectively . here ,
the compositions reduce to the recursive transformation @xmath466 with @xmath467 being equal to @xmath468 or @xmath469 and @xmath384 . the left - hand - side of expression ( 66 )
can again be presented at each @xmath145 as a single exponential , @xmath470 where now @xmath471 { \cal y}_{k+3 } & = & q_3 { \cal x}_{k+3 } + q_4 [ { \cal x}_1,{\cal x}_1,{\cal x}_{k+1 } ] \ , , \nonumber \\ [ 6pt ] { \cal y}_{k+5 } & = & q_5 { \cal x}_{k+5 } + q_6 [ { \cal x}_1,{\cal x}_1,{\cal x}_{k+3 } ] + \nonumber \\ [ -6pt ] \\ [ -6pt ] & & q_7 [ { \cal x}_1,{\cal x}_1,{\cal x}_1,{\cal x}_1,{\cal x}_{k+1 } ] , \nonumber \\ [ 6pt ] { \cal y}_{k+7 } & = & q_8 { \cal x}_{k+7 } + q_9 [ { \cal x}_1,{\cal x}_1,{\cal x}_{k+5 } ] + \nonumber \\ [ 1pt ] & & q_{10 } [ { \cal x}_1,{\cal x}_1,{\cal x}_1,{\cal x}_1,{\cal x}_{k+3 } ] + \nonumber \\ [ 1pt ] & & q_{11 } [ { \cal x}_1,{\cal x}_1,{\cal x}_1,{\cal x}_1 , { \cal x}_1,{\cal x}_1,{\cal x}_{k+1 } ] \ , . \nonumber\end{aligned}\ ] ] recursive relations for multipliers @xmath472 , corresponding to transformation ( 66 ) , remain the same in form as in the case @xmath59 . so that we should merely to put either @xmath72 or @xmath74 in eq .
( 60 ) to obtain the required set of order conditions . in view of the equivalence of eqs .
( 54 ) and ( 55 ) at @xmath473 , the first step on increasing the order of composition scheme to @xmath428 when @xmath72 or @xmath443 when @xmath74 is trivial and yields @xmath61 , @xmath474 , and @xmath426 .
the next steps on increasing @xmath369 to the higher values @xmath475 , @xmath476 , and @xmath477 at @xmath72 or 8 are similar to the steps described above for @xmath59 .
namely , they lead to the necessity of solving numerically the system of @xmath78 non - linear equations , @xmath419 , @xmath420 , , @xmath478 , with @xmath66 , 7 , and 11 , respectively .
the only difference from the case @xmath59 is that at @xmath72 or 8 and @xmath479 , the number of equations reduces from 12 to @xmath480 , because of a somewhat simplified structure of the last truncation operator shown in eq .
( 67 ) with respect to that appearing in eq .
( 59 ) . so that below we will present final results only with brief comments for each the above cases .
the best set among the solutions found were identified as those that minimize the quantity @xmath481 ( almost always this led to the minimization of the norm for the main term of truncation errors as well ) . for @xmath72 and @xmath443
there are five solutions with the best set @xmath482 d_2 = \ \ 0&.&11922404430206648052593264029266{\rm e}\!+\!00 \\ [ -2pt ] d_3 = -0&.&10677277516805770678518370004925{\rm e}\!+\!01 \nonumber\end{aligned}\ ] ] with @xmath483 and @xmath484 ( within three significant digits ) . at @xmath74 and @xmath456
we have found again five solutions and the optimal one is @xmath485 d_2 = \
\ 0&.95777180465215511634906238400062&{\rm e}\!-\!01 \\ [ -2pt ] d_3 = -0&.10545412798113627599734519738778&{\rm e}\!+\!01 \nonumber\end{aligned}\ ] ] with @xmath483 and @xmath486 . for @xmath72 and @xmath456
there were more than 150 solutions with the optimal set @xmath487 d_2 = \ \
0&.&44631941526959576960102601257986 \nonumber \\ [ -2pt ] d_3 = -0&.&66447133641046221008529452937721 \nonumber \\ [ -8pt ] \\ [ -8pt ] d_4 = -0&.&58260619571844248816548809046510 \nonumber \\ [ -2pt ] d_5 = \
\ 0&.&64081619589013117205634311707157 \nonumber \\ [ -2pt ] d_6 = \ \ 0&.&31805596598883340430918587031701 \nonumber\end{aligned}\ ] ] and @xmath449 with @xmath488 .
when @xmath74 and @xmath489 we have computed more than 150 solutions also and identified among them the following optimal one @xmath490 d_2 = \ \
0&.&46763050598682150405078600842681 \nonumber \\ [ -2pt ] d_3 = -0&.&63245030403272077359889720182431 \nonumber \\ [ -8pt ] \\ [ -8pt ] d_4 = -0&.&58223379020720528275072356442667 \nonumber \\ [ -2pt ] d_5 = \
\ 0&.&62109852451075548059651686410928 \nonumber \\ [ -2pt ] d_6 = \ \ 0&.&29686555238409826518407483052733 \nonumber\end{aligned}\ ] ] with @xmath449 and @xmath491 . at @xmath72 and
@xmath489 the best set , among more than 200 solutions realized , is @xmath492 d_2 = -0&.&47389771786834222637653653795835 \nonumber \\ [ -2pt ] d_3 = \ \
0&.&54376649763596364670254533524499 \nonumber \\ [ -2pt ] d_4 = -0&.&64055411141298491334240825973418 \nonumber \\ [ -2pt ] d_5 = \
\ 0&.&28139025047030322588052971757542 \nonumber \\ [ -8pt ] \\ [ -8pt ] d_6 = \
\ 0&.&56345778618405675650229011409013 \nonumber \\ [ -2pt ] d_7 = \
\ 0&.&64205004597526944181678051477448 \nonumber \\ [ -2pt ] d_8 = -0&.&16972825772391310721875128881451 \nonumber \\ [ -2pt ] d_9 = -0&.&57973031669054683392549871514985 \nonumber \\ [ -2pt ] d_{10 } = \ \ 0&.&27398580283063379870623390979762 \nonumber\end{aligned}\ ] ] with @xmath493 and @xmath494 . finally for @xmath74 and @xmath495 the optimal solution , among again more than 200 sets calculated , is @xmath496 d_2 = \ \
0&.&55268563185301488324882994018746 \nonumber \\ [ -2pt ] d_3 = -0&.&58134339535533393315605544309940 \nonumber \\ [ -2pt ] d_4 = \
\ 0&.&23403665265420481243563202333267 \nonumber \\ [ -2pt ] d_5 = -0&.&51788958989817055303978658827453 \nonumber \\ [ -8pt ] \\ [ -8pt ] d_6 = -0&.&43983975477992920522811970527874 \nonumber \\ [ -2pt ] d_7 = -0&.&20137078150942169957468111993444 \nonumber \\ [ -2pt ] d_8 = \
\ 0&.&34412872002528894622975927197416 \nonumber \\ [ -2pt ] d_9 = \
\ 0&.&03072591760996558798895428309765 \nonumber \\ [ -2pt ] d_{10 } = \ \ 0&.&48652953960727041281280535031455 \nonumber\end{aligned}\ ] ] with @xmath493 and @xmath497 decreases with increasing @xmath369 at any @xmath2 ( 4 , 6 , and 8) considered .
besides the improvement of the precision of the integration , this leads to an extension of the region of stability of the composition algorithms . indeed
, the constants @xmath368 appear in the compositions ( see eq .
( 54 ) ) in the form of the term @xmath498 ( and its combinations of different orders when evaluating truncation uncertainties @xmath499 ) .
then , taking into account that @xmath481 , the maximal value for the size @xmath25 of the time step , at which these uncertainties do not exceed an acceptable level of precision , can be estimated as @xmath500 .
this also explains a well correlation of @xmath501 with the minimum for the norm of truncation errors .
sixth- and eighth - order based compositions may have advantages over algorithms basing on fourth - order schemes especially when constructing very precise integrators with high values of @xmath369 .
for instance , in order to derive an integrator of order @xmath495 on the basis of triplet concatenation ( 55 ) of a scheme of order @xmath65 , it is necessary to apply @xmath502 fourth - order stages .
taking into account that each such a stage requires @xmath503 force and @xmath260 force - gradient evaluations ( see eqs . ( 38 ) or ( 39 ) ) , one obtains in total the numbers @xmath504 and @xmath505 corresponding to a whole time step . on the other hand , in view of result ( 73 ) , an integrator of order @xmath495 can be composed at @xmath74 and @xmath480 using @xmath506 eighth - order stages for each of which @xmath507 ( see eq .
( 53 ) ) . so that the overall number of force and force - gradient recalculations will be equal only to 231 that is much smaller than the above values 2187 and 729 obtained in the case @xmath65
in molecular dynamics ( md ) simulations we dealt with a system of @xmath508 identical ( @xmath509 ) particles interacting through a lennard - jones potential , @xmath510 $ ] .
the particles were placed in a cubic box of volume @xmath511 and periodic boundary conditions have been used to exclude the finite - size effects . for the same reason ,
the initial potential was modified as @xmath512 at @xmath513 with @xmath514 for @xmath515 , where @xmath516 is the cut - off radius .
then the potential @xmath517 and its first - order derivative @xmath518 will be continuous functions any where in @xmath519 including the truncation point @xmath520 .
this avoids an energy drift caused by the passage of particles via the surface of truncation sphere as well as singularities of @xmath521 and @xmath522 at @xmath520 .
the simulations were carried out in a microcanonical ensemble at a reduced density of @xmath523 and a reduced temperature of @xmath524 .
all runs of the length in @xmath525 time steps each were started from an identical well equilibrated initial configuration @xmath526 .
the precision of the integration was measured in terms of the relative total energy fluctuations @xmath527 , where @xmath528 and @xmath529 denotes the microcanonical averaging .
the equations of motion were integrated using force - gradient algorithms ( 30 ) , ( 36 ) , and ( 38 ) of the fourth order within schemes a , a@xmath257 , b , c , and c@xmath257 ( see eqs . (
32 ) , ( 33 ) , ( 37 ) , ( 41 ) , and ( 42 ) , respectively ) . for the purpose of comparison the integration with the help of usual fourth - order algorithm by forest and ruth ( fr ) @xcite ( which represents , in fact
, triplet concatenation ( 55 ) of second - order verlet scheme ( 8) ) has been performed as well .
the corresponding results for the total energy fluctuations as functions of the length @xmath26 of the simulations is presented in subset ( a ) of fig . 1 for a typical reduced time step of @xmath530 .
as can be seen , schemes a , b , and c exhibit a similar equivalence in energy conservation .
this is in agreement with our theoretical predictions presented in subsection ii.b._4 _ , where the precision of algorithms has been estimated in terms of the norm @xmath244 ( eq . ( 34 ) ) of fifth - order truncation errors @xmath236 . in particular for schemes a , b , and c , it has been obtained @xmath531 , 0.000715 , and 0.000715 , respectively .
further , as expected , scheme a@xmath257 ( @xmath532 ) is worse in precision and leads to values of @xmath533 which are approximately in @xmath534 times larger .
note that in microcanonical ensembles the total energy is an integral of motion , @xmath535 , so that within approximate md simulations , smaller values of @xmath533 correspond to a better precision of the integration .
it is worth remarking also that another integral of motion , namely , total momentum @xmath536 , is conserved exactly within force - gradient approach ( 13 ) .
the reasons are that all velocities are updated at once ( see eq .
14 ) during each stage of decompositions and the fact that @xmath537 as well as @xmath538 ( as can be verified readily using the structure of eq . ( 11 ) ) . the best accuracy in energy conservation can be achieved within optimized scheme c@xmath257 ( see fig . 1 ( a ) ) for which @xmath539 .
it minimizes @xmath533 to a level of @xmath540 that is in factor @xmath541 lower than those related to schemes a , b , and c. at the same time , the usual fr algorithm leads to the worst result @xmath542 .
we see , therefore , that applying force - gradient approach allows to reduce unphysical energy fluctuations up in two orders in magnitude .
let us show now that this overcompensates an increased computational efforts caused by additional calculations of the force gradients . the processor time used for carrying out these calculations ( see eq .
( 11 ) ) was nearly in 3 times larger than that needed for evaluations of forces itself .
further , we should take into account that algorithm c@xmath257 requires @xmath503 force and @xmath260 force - gradient recalculations per time step , whereas @xmath503 and @xmath543 for fr scheme . as a result ,
one obtains that the size @xmath25 of one step within fr propagation must be in @xmath544 times shorter than in the case of algorithm c@xmath257 for spending the same overall processor time within both the cases during the integration over a fixed time interval . finally , in view of the fact that the global error and thus the function @xmath533 are proportional to the fourth power of @xmath25 , i.e. , @xmath545 , one finds that , at the above conditions , the level of conservation of the fr scheme reduces from @xmath546 to @xmath547 . so that relative efficiency of scheme c@xmath257 with respect to the fr integrator is actually equal to @xmath548 . in order to ensure that scheme c@xmath257 ( eq . ( 42 ) ) is indeed the best among whole family ( 40 ) of c@xmath257-like integrators ( 38 )
, we carried out additional simulations in which the parameter @xmath228 , being constant within each simulation , varied from one run to another .
the total energy fluctuations obtained in such simulations at the end of the runs for two ( fixed within each run ) undimensional time steps , namely , @xmath549 and 0.005 , are shown in subset ( b ) of fig . 1 as functions of @xmath228 .
as can be observed , the dependencies @xmath550 have the global minimum located at the same point @xmath551 independently on the size @xmath25 of the time step .
this point coincides completely with the minimum given by eq .
( 42 ) for the function @xmath268 ( see eq . ( 34 ) and relations following just after eq .
( 41 ) ) which is also included in the subset and plotted by a dashed curve ( where an upper lying part of the curve corresponds to sign plus in eq .
( 40 ) , whereas a lower lying part as well as the simulation data are related to sign minus ) .
so that our criterion on measuring the precision of the integration in terms of the norm of local truncated uncertainties is in excellent accord .
moreover , the energy fluctuations appear to be proportional to that norm @xmath244 as @xmath552 and the coefficient of the resulting proportionality almost does not depend on @xmath228 and @xmath25 . 2 it is worth remarking that the results reported in this subsection should be considered as the first attempts of applying force - gradient algorithms to md simulations .
in previous studies , algorithms of such a kind have been tested for classical @xcite and quantum @xcite mechanics systems composed of a few bodies only ( or even one body moving in an external field ) .
the present investigations have demonstrated that force - gradient algorithms can be used with equal success in statistical mechanics simulations dealing with a great number of particles , i.e. , when @xmath553 . in the last case
, the calculations of force gradients also presents no difficulties . indeed , during the integration we should first evaluate usual forces @xmath554 for each particle @xmath6 , where @xmath555 .
this involves a number of operations which is proportional to the second power of @xmath3 . then in view of the structure of eq . ( 11 ) and taking into account the fact that particle s accelerations @xmath556 are already known quantities , the calculations of gradients @xmath557 will require a number of operations which is proportional to the same power of @xmath3 , i.e. , @xmath558 ( but not to @xmath559 , as it may look at first sight ) . further reducing the computational efforts
is possible taking into account that function @xmath560 ( see eq .
( 11 ) ) decreases with increasing the interparticle separation @xmath561 more rapidly than initial potential @xmath562 .
in such a case , a secondary cut - off radius @xmath563 can be introduced when evaluating @xmath564 ( that is equivalent to putting @xmath565 at @xmath566 ) in order to speed up the calculations .
one of the simplest way to test force - gradient algorithms of higher orders is to apply them to solution of the two - dimensional kepler problem . in particular , this way has been chosen by chin and kidwell @xcite when testing fourth - order algorithms a , b , and c and higher - order iterated counterparts of the last scheme . as has been established , this force - gradient scheme is particularly outstanding and appears to be much more superior than usual non - gradient integrators , such as fourth - order by forest and ruth @xcite as well as by runge and kutta @xcite , sixth - order by yoshida @xcite , etc . in this subsection
it will demonstrated that further significant improvement in the effectiveness of the integration can be reached replacing standard iteration procedure ( 55 ) by advanced composition approach ( 54 ) .
moreover , using our new sixth- and eighth - order force - gradient algorithms as the basis for the composition has allowed us to perform the computations with extremely high precision which exceeds by several orders the accuracy observing within standard fourth - order based schemes .
we will consider a motion of a particle ( planet ) of mass @xmath567 moving in the ( gravitation ) field @xmath568 of the central body ( sun ) with mass @xmath569 , where @xmath570 is the constant responsible for intensity of the interaction . for simplifying the calculations ,
one neglects the influence of all other ( @xmath571 ) particles ( planets , for which @xmath572 ) in the ( solar ) system .
then the motion can be described by the following system of two equations , @xmath573 where @xmath574 , and for clarity of the presentation we have used units in which the reduced mass @xmath575 and the interaction constant @xmath576 are equal to unity . since the quantity @xmath577 ( which is associated with the total energy ) presents an integral of motion for equations ( 74 ) , it should be conserved during the integration .
however , this will so if these equations are solved exactly . in numerical simulations ,
the local truncation uncertainties @xmath499 accumulate step by step of the integration process , leading at @xmath578 to the global errors @xmath579 , where @xmath369 denotes the order of a self - adjoint algorithm . so that the quantity @xmath580 can be presented as a function of time as @xmath581 where @xmath582 and @xmath583 is the main step - size independent error coefficient .
in our simulations we solved two - dimensional kepler problem ( 74 ) with the same initial conditions @xmath584 and @xmath585 as those used by previous authors @xcite to make comparative analysis more convenient .
the resulting highly eccentric ( @xmath586 ) orbit provides a nontrivial testing ground for trajectory integration .
the numerical effectiveness of each algorithm was gauged in terms of main error coefficient @xmath587/\delta t^q$ ] ( see eq . ( 75 ) ) .
it can actually be extracted from the fraction @xmath588/\delta t^q$ ] by choosing smaller and smaller time steps @xmath25 to be entitled to completely ignore next higher - order corrections @xmath589 .
this typically occurs in the neighborhood of @xmath590 , where @xmath591 is the period of the elliptical orbit . since we are dealing with algorithms of high orders @xmath369 and
small step sizes @xmath25 , all the calculations have been carried out in fortran using quadruple ( instead of double , as for md simulations ) precision arithmetics for ensuring the correctness of the results .
the normalized energy deviations @xmath592 obtained in the simulations applying fourth- , sixth- , eighth- , tenth- , twelfth- , and fourteenth - order algorithms are plotted in subsets ( a ) , ( b ) , ( c ) , ( d ) , ( e ) , and ( f ) of fig . 2 , respectively , as functions of time @xmath593 during one period @xmath78 of the orbit .
these deviations are substantial only near mid period when the particle is at its closest position to the attractive center .
note also that within symplectic integration , the nonconservation of energy for periodic orbits is periodic and its averaged ( over times @xmath594 ) value is bounded and independent on @xmath593 ( the independence of averaged energy fluctuations at @xmath578 has already been demonstrated in md simulations , see fig . 1 ) .
that is way we presented the results in fig . 2 within a narrow region of time near @xmath595 , where the maximal deviations of @xmath583 will give a main contribution to the overall fluctuations .
in the case of fourth - order integration we used most typical algorithms a , b , c , and c@xmath257 ( see eqs .
( 32 ) , ( 37 ) , ( 41 ) , and ( 42 ) , respectively ) . as can be seen from subset ( a ) of fig . 2
, the pattern here is somewhat different than that in md simulations ( please compare with fig . 1 ( a ) ) .
the algorithm c is clearly better than schemes a and b , that confirms the conclusion of ref . @xcite . on the other hand , integrator c@xmath257 does not exhibit an improved precision in energy conservation with respect to scheme c. nearly the same was seen when iterating these algorithms to higher orders with the help of triplet construction ( 55 ) . in particular , the sixth - order c@xmath257-counterpart appeared to be even slightly better than the corresponding counterpart of scheme c ( see subset ( b ) of fig . 2 ) .
at the same time , higher - order integrators basing on schemes a and b were definitely worse . so that the obvious candidates for fourth - order based iterations ( 55 ) and compositions ( 54 ) are schemes c and c@xmath257 .
in order to understand why scheme c@xmath257 does not lead to the expected improvement over scheme c in this particular situation , it should be taken into account that we deal with a small system , actually with one body moving in an effective external field .
moreover , such a body moves periodically and , thus , covers only small part of phase space during its displacement .
this is contrary to many - body statistical systems , where the phase point may visit considerably wider regions of phase space . in the latter case ,
during the averaging along the phase trajectories , different components @xmath131 of fifth - order local uncertainties ( see eq . ( 17 ) ) will enter with approximately the same weights when forming the total error vector @xmath236 .
this has been tentatively assumed when writing the norm @xmath244 of that vector in the form of eq .
( 34 ) and further minimizing @xmath244 to obtain algorithm c@xmath257 . in the case of a few - body system , especially with periodic motion , the above weights may differ considerably .
this complicates an analysis of the truncation terms and makes impossible to find an exact global minimum for them within any analytical approach .
note , however , that even here , the assumption on uniform contribution of truncation - error components works relatively well . indeed , in view of dependencies shown in subsets ( a ) and ( b )
, we can stay that both the schemes c and c@xmath257 are comparable in precision . the same was observed for their higher - order counterparts .
for this reason ( and to reserve more free space for other dependencies ) , in further subsets ( c)(f ) we will draw only curves corresponding to scheme c. 2 when considering the sixth - order integration , we realized that direct velocity - like scheme defined by eqs .
( 45 ) and ( 47 ) is much worse ( the maximum deviation of @xmath583 were more than in two orders larger ) than its extended position - like counterpart given by eqs .
( 50 ) and ( 51 ) .
this is in agreement with a prediction of subsection ii.b._5_. the result corresponding to the position - like algorithm is plotted in subset ( b ) of fig . 1 by the bold dashed curve marked as g6 .
as can be seen , all three curves shown in this subset , namely c , c@xmath257 , and g6 are close enough to each other . but algorithm g6 uses only @xmath299 force evaluations per time step , instead of @xmath596 needed for iterated c- and c@xmath257-like schemes ( for all these three cases the number of force - gradient evaluations is the same and equal to @xmath300 ) . therefore , for order six , direct decomposition approach ( 13 ) leads to more efficient integration than concatenations of fourth - order schemes .
beginning from order eight , the above concatenations based on standard iterations ( 55 ) and advanced compositions ( 54 ) will result in completely different integrators .
the simulation data for these iterated and composed c - based integrators are shown in subset ( c ) of fig . 2 by thin ( marked simply as c ) and bold ( marked as s ) solid curves , respectively .
the curves related to tenth- and twelfth - order iteration and composition integrators ( based on the same fourth - order scheme c ) are plotted correspondingly in subsets ( d ) and ( e ) of fig . 2 , and marked by the same letters c and s. we mention that c - marked curves have already been presented in the work by chin and kidwell @xcite up to order 12 .
they are redisplayed by us in order to illustrate the evident superiority of our new composition approach over the standard iteration method . indeed , for the iteration integrators ( c - marked curves ) of orders @xmath428 , 10 , and 12 , the magnitudes of the normalized energy coefficient @xmath592 after one period are 1.44 , 19.24 , and 424.8 , respectively .
on the other hand , the magnitudes related to the composition integrators ( s - marked curves ) constitute correspondingly 0.0953 , 0.0577 , and 1.41 , i.e. , they are approximately in 15 , 330 , and 300 times smaller . in addition , the composition integrators are faster with respect to their iteration versions in factors 9/7 , 27/13 , and 81/23 for @xmath428 , 10 , and 12 , respectively ( see subsection iii.a._1 _ ) , and thus the resulting efficiencies will increase yet . what about sixth- and eight - order - based composition schemes at @xmath597 ?
first of all , let us consider the case of eight - order integration . here , the direct scheme chosen was position - like integrator ( 53 ) ( it leads to better energy conservation with respect to its velocity - like counterpart ( 52 ) ) .
the result corresponding to this integrator is plotted in subset ( c ) of fig . 2 by the dashed curve marked as g8 .
as can be seen , the fourth - order - based composition scheme ( s - curve ) is better at @xmath428 with respect to both direct g8 and iterated g6-like versions . with increasing the order to 10 and 12 , all they become to be nearly equivalent in the accuracy of energy conservation .
but fourth - order - based approach requires somewhat fewer number of operations .
for instance , for order 12 , one obtains that the numbers of force and gradients evaluations per time step are equal for it to @xmath598 and @xmath599 , respectively , whereas these numbers for sixth- and eighth - order - based compositions g6 and g8 are @xmath600 , @xmath601 and @xmath602 ( where g6-integrator requires less operations than g8-scheme ) . however , beginning from order 14 , the situation reverses .
the fourth - order - based composition s - approach becomes to be not longer accessible ( because of the absence of explicit expressions for its time coefficients here ) . on the other hand , applying the standard fourth - order - based iteration c - method is very inefficient .
in particular , at @xmath489 the maximal energy deviations within this method is @xmath603 with @xmath604 and @xmath605 . at the same time , the higher - order - based composition schemes lead to much accurate results , namely , @xmath606 with @xmath607 and @xmath608 for g6- as well as @xmath609 with @xmath610 for g8-based schemes ( where the better precision for the last scheme compensates to some extent its increased values for quantities @xmath295 and @xmath297 ) .
we see , therefore , that the relative efficiencies of g6- and g8-based schemes with respect to c - approach constitute about @xmath611@xmath612 .
finally , in the case @xmath495 ( not shown in fig . 2 ) we have obtained the values @xmath613 and @xmath614 corresponding to c- and g8-based schemes , respectively .
taking into account the numbers of @xmath295 and @xmath297 for these schemes presented at the end of subsection ii.c._2 _ , one can conclude that the efficiency increases here also approximately in @xmath611 to @xmath612 times .
in this work we have formulated a general theory of construction of force - gradient algorithms for solving the equations of motion in classical and quantum systems .
this has allowed us to extend considerably the class of analytically integrable symplectic schemes .
the new algorithms derived include self - adjoint direct decomposition integrators of orders two , four , six , and eight as well as their composition counterparts up to the sixteenth order in the time step .
as has been proven theoretically and confirmed in actual numerical simulations , these algorithms lead to significant improvement in the efficiency of the integration with respect to existing force - gradient and non - gradient schemes .
it has been demonstrated that force - gradient algorithms can be used with equal success as for describing the motion in few - body classical and quantum mechanics systems as well as for performing statistical molecular dynamics observations over many - particle collections . in all the cases
the calculation of force gradients presents no difficulties and requires computational efforts comparable with those needed to evaluate usual forces itself
. the new algorithms may be especially useful for the prediction and study of very subtle dynamical effects in different areas of physics and chemistry including the problems of astrophysical interest , whenever the precise integration of motion during very long times is desirable .
the algorithms introduced exactly reproduce such important features of classical dynamics as time reversibility and symplecticity .
this explains their excellent energy conservation and stability properties . in this context
it should be mentioned another class of ( non - gradient ) integrators recently developed @xcite on the basis of a modified runge - kutta approach . like the force - gradient algorithms ,
the runge - kutta - like integrators also allow to produce time - reversible and symplectic trajectories in phase space with , in principle , arbitrary order in precision .
however , such integrators are implicit and require cumbersome systems of globally coupled ( via positions and forces of all particles ) nonlinear equations be solved by expensive iterations at each step of the integration process . since in practice such equations can not be solved exactly , the time reversibility and symplecticity can be violated .
this may lead , in particular , to instabilities in long - term energy conservation , i.e. , to the same problem inherent in the tradition ( nonsymplectic ) runge - kutta method ( see the introduction ) .
all these disadvantages are absent in the present approach , where the phase trajectories are propagated explicitly in time by applying consecutive simple shifts of particles in velocity and position space with exact preservation of the phase volume and reversibility of the generated solutions .
the approach presented can also be adapted to the integration of motion in more complicated systems , such as systems with orientational or spin degrees of freedom , etc . , where splitting of the liouville operator into more than two parts may be necessary to obtain analytically solvable subpropagators .
these and other related problems will be considered in a separate investigation .
part of this work was supported by the fonds zur frderung der wissenschaftlichen forschung under project no . 15247 . the recursive relations for the highest - order multipliers @xmath132 ( see eqs .
( 13 ) , ( 15 ) , and ( 18 ) ) corresponding to the first type of self - adjoint transformations given by the first line of eq .
19 are : @xmath620 \zeta_6^{(n+1 ) } = \zeta_6^{(n ) } + & & a^{(n ) } \big(18 ( { a^{(n)}}^3 { \sigma^{(n)}}^3 - 84 \alpha^{(n ) } \big(15 \beta^{(n ) } - \big(a^{(n ) } + \nu^{(n)}\big ) { \sigma^{(n)}}^2\big ) + ( { a^{(n)}}^2 \big ( 15 \nu^{(n ) } { \sigma^{(n)}}^3 - 252 \beta^{(n ) } \sigma^{(n ) } \big ) + \\ & & 6 a^{(n ) } \big(210 \gamma_4^{(n ) } - 28 \beta^{(n ) } \nu^{(n ) } \sigma^{(n ) } + { \nu^{(n)}}^2 { \sigma^{(n)}}^3\big ) + 2 \big(630 \gamma_4^{(n ) } \nu^{(n ) } - \\ & & 630 \gamma_2^{(n ) } \sigma^{(n ) } - 42 \beta^{(n ) } { \nu^{(n)}}^2 \sigma^{(n ) } + { \nu^{(n)}}^3 { \sigma^{(n)}}^3\big)\big)/7560 \ , , \end{aligned}\ ] ] @xmath629 \zeta_5^{(n+1 ) } = \zeta_5^{(n ) } - & & \big ( 2520 { \alpha^{(n)}}^2 b^{(n ) } - 36 { b^{(n)}}^3 { \nu^{(n)}}^4 + 42 c^{(n ) } { \nu^{(n)}}^3 \sigma^{(n ) } + 30 { b^{(n)}}^2 { \nu^{(n)}}^4 \sigma^{(n ) } + 168 \alpha^{(n ) } \nu^{(n ) } \big ( 15 c^{(n ) } - \\ & & b^{(n ) } \nu^{(n ) } \big(6 b^{(n ) } + \sigma^{(n)}\big)\big ) - b^{(n ) } \big(15120 \gamma_3^{(n ) } \nu^{(n ) } - { \nu^{(n)}}^3 \big(252 \beta^{(n ) } + 504 c^{(n ) } + \nu^{(n ) } { \sigma^{(n)}}^2\big)\big)\big)/45360 \ , , \end{aligned}\ ] ] all these relations as well as other symbolic expressions presented in the work has been carried out using mathematica 4.0 and maple 6 packages installed on the silicon graphics origin 3800 workstation at linz university .
the numerical calculations have also been performed there . | a consequent approach is proposed to construct symplectic force - gradient algorithms of arbitrarily high orders in the time step for precise integration of motion in classical and quantum mechanics simulations . within this approach the basic algorithms
are first derived up to the eighth order by direct decompositions of exponential propagators and further collected using an advanced composition scheme to obtain the algorithms of higher orders .
contrary to the scheme by chin and kidwell [ phys .
rev .
e * 62 * , 8746 ( 2000 ) ] , where high - order algorithms are introduced by standard iterations of a force - gradient integrator of order four , the present method allows to reduce the total number of expensive force and its gradient evaluations to a minimum . at the same time
, the precision of the integration increases significantly , especially with increasing the order of the generated schemes .
the algorithms are tested in molecular dynamics and celestial mechanics simulations .
it is shown , in particular , that the efficiency of the new fourth - order - based algorithms is better approximately in factors 5 to 1000 for orders 4 to 12 , respectively .
the results corresponding to sixth- and eighth - order - based composition schemes are also presented up to the sixteenth order . for
orders 14 and 16 , such highly precise schemes , at considerably smaller computational costs , allow to reduce unphysical deviations in the total energy up in @xmath0 times with respect to those of the standard fourth - order - based iteration approach .
pacs numbers : 02.60.cb ; 05.10.-a ; 95.10.ce ; 95.75.pq 2 |
we wish to thank the kekb accelerator group for the excellent operation of the kekb accelerator .
we acknowledge support from the ministry of education , culture , sports , science , and technology of japan and the japan society for the promotion of science ; the australian research council and the australian department of industry , science and resources ; the national science foundation of china under contract no .
10175071 ; the department of science and technology of india ; the bk21 program of the ministry of education of korea and the chep src program of the korea science and engineering foundation ; the polish state committee for scientific research under contract no .
2p03b 17017 ; the ministry of science and technology of the russian federation ; the ministry of education , science and sport of the republic of slovenia ; the national science council and the ministry of education of taiwan ; and the u.s .
department of energy .
the fox - wolfram moments were introduced in g. c. fox and s. wolfram , phys .
lett . * 41 * , 1581 ( 1978 ) .
the fisher discriminant used by belle is described in k. abe _ et al . _
( belle collab . )
lett . * 87 * , 101801 ( 2001 ) and k. abe _ et al . _
( belle collab . ) , phys . lett . *
b511 * , 151 ( 2001 ) . | we report results on the decay @xmath0 and its charge conjugate using a data sample of 85.4 million @xmath1 pairs recorded at the @xmath2 resonance with the belle detector at the kekb asymmetric @xmath3 storage ring
. ratios of branching fractions of cabibbo - suppressed to cabibbo - favored processes are determined to be @xmath4 , @xmath5 and @xmath6 where the indices 1 and 2 represent the cp=+1 and cp=@xmath71 eigenstates of the @xmath8 system , respectively .
we find the partial - rate charge asymmetries for @xmath0 to be @xmath9 and @xmath10 .
the extraction of @xmath11 @xcite , an angle of the kobayashi - maskawa triangle @xcite , is a challenging measurement even with modern high luminosity @xmath12 factories .
recent theoretical work on @xmath12 meson dynamics has demonstrated the direct accessibility of @xmath11 using the process @xmath13
@xcite . if the @xmath14 is reconstructed as a cp eigenstate , the @xmath15 and @xmath16 processes interfere .
this interference leads to direct cp violation as well as a characteristic pattern of branching fractions .
however , the branching fractions for @xmath17 meson decay modes to cp eigenstates are only of order 1 % . since cp violation through interference
is expected to be small , a large number of @xmath12 decays is needed to extract @xmath18 . assuming the absence of @xmath19 mixing , the observables sensitive to cp violation that are used to extract the angle @xmath18 @xcite are , @xmath20 where the ratios @xmath21 and @xmath22 are defined as @xmath23@xmath24@xmath25 and @xmath26 are cp - even and cp - odd eigenstates of the neutral @xmath17 meson , @xmath27 is the ratio of the amplitudes of the two tree diagrams shown in fig . 1 and @xmath28 is their strong - phase difference
. the ratio @xmath29 corresponds to the magnitude of cp asymmetry and is suppressed to the level of @xmath30 due to the ckm factor @xmath31 and a color suppression factor @xmath32 .
note that the asymmetries @xmath33 and @xmath34 have opposite signs .
the ratio of the cabibbo - suppressed decay @xmath35 to the cabibbo - favored decay @xmath36 has been reported by cleo @xcite to be @xmath37 while belle finds @xmath38 @xcite .
assuming factorization , the ratio @xmath22 is expected to be @xmath39 in the tree - level approximation , where @xmath40 is the cabibbo angle , and @xmath41 and @xmath42 are meson decay constants .
the measurements are in good agreement with this theoretical expectation . previously , belle reported the observation of the decays @xmath43 and @xmath44 with @xmath45 @xcite . this paper reports more precise measurements of these decays with a data sample of @xmath46 , containing 85.4 million @xmath1 pairs , collected with the belle detector at the kekb asymmetric - energy @xmath47 ( 3.5 on 8 gev ) collider operating at the @xmath2 resonance . at kekb , the @xmath2
is produced with a lorentz boost of @xmath48 nearly along the electron beamline .
the belle detector is a large - solid - angle magnetic spectrometer that consists of a three - layer silicon vertex detector ( svd ) , a 50-layer central drift chamber ( cdc ) , an array of silica aerogel threshold erenkov counters ( acc ) , a barrel - like arrangement of time - of - flight scintillation counters ( tof ) , and an electromagnetic calorimeter ( ecl ) comprised of csi(tl ) crystals located inside a super - conducting solenoid coil that provides a 1.5 t magnetic field .
an iron flux - return located outside of the coil is instrumented to detect @xmath49 mesons and to identify muons ( klm ) .
the detector is described in detail elsewhere @xcite .
we reconstruct @xmath14 mesons in the following decay channels . for the flavor specific mode ( denoted by @xmath50 ) , we use @xmath51 @xcite . for cp
= + 1 modes , we use @xmath52 and @xmath53 while for cp = @xmath71 modes , we use @xmath54 , @xmath55 , @xmath56 , @xmath57 and @xmath58 .
the charged track , @xmath59 and @xmath60 selection requirements have been described in ref . @xcite . for each charged track , information from the acc , tof and specific ionization measurements from the cdc are used to determine a @xmath61 likelihood ratio @xmath62 = @xmath63 , where @xmath64 and @xmath65 are kaon and pion likelihoods . for kaons ( pions ) from the @xmath66 mode we used the particle identification requirement of @xmath67 . for kaons from the @xmath68 mode
we require @xmath69 while for pions from @xmath70 mode we require @xmath71 .
the @xmath72 mesons are reconstructed from @xmath73 combinations in the mass window @xmath74 with the charged pion particle identification requirement @xmath75 .
to reduce the contribution from the non - resonant background , a helicity angle cut @xmath76 is applied where @xmath77 is the angle between the normal to the @xmath72 decay plane in the @xmath72 rest frame and the @xmath72 momentum in the @xmath14 rest frame . to remove the contribution from @xmath78
, we require the @xmath79 invariant mass to be greater than @xmath80 from the @xmath81 nominal mass .
the @xmath82 mesons are reconstructed from two oppositely charged kaons in the mass window of @xmath83 with @xmath84 .
we also apply the @xmath82 helicity angle cut @xmath76 where @xmath77 is the angle between one of the @xmath82 daughters in the @xmath82 rest frame and the @xmath82 momentum in the @xmath14 rest frame .
we form candidate @xmath85 and @xmath86 mesons using the @xmath87 and @xmath88 decay modes with mass ranges of @xmath89 and @xmath90 , respectively .
the @xmath85 momentum is required to be greater than @xmath91 .
both @xmath85 and @xmath86 candidates are kinematically constrained to their nominal masses .
the @xmath14 candidates are required to have masses within @xmath922.5@xmath93 of their nominal masses , where @xmath93 is the measured mass resolution which ranges from @xmath94 to @xmath95 depending on the decay channel . a @xmath14 mass and ( wherever possible)vertex constrained fit is then performed on the remaining candidates .
we combine the @xmath14 and @xmath96/@xmath97 candidates ( denoted by @xmath98 ) to form @xmath12 candidates . we apply tighter particle identification cuts , @xmath99 for prompt kaons ( pions ) , to identify @xmath100 events .
the signal is identified by two kinematic variables calculated in the center - of - mass ( c.m . )
frame .
the first is the beam - energy constrained mass , @xmath101 , where @xmath102 and @xmath103 are the momenta of @xmath14 and @xmath104 candidates and @xmath105 is the beam energy in the c.m
. frame .
the second is the energy difference , @xmath106 , where @xmath107 is the energy of the @xmath14 candidate , @xmath108 is the energy of the @xmath104 candidate calculated from the measured momentum and assuming the pion mass,@xmath109 . with this definition , real @xmath110 events peak at @xmath111 even when they are misidentified as @xmath35 , while @xmath112 events peak around @xmath113 mev .
event candidates are accepted if they have @xmath114 and @xmath115 . in case of multiple candidates from a single event
, we choose the best candidate on the basis of a @xmath116 determined from the differences between the measured and nominal values of @xmath117 and @xmath118 . to suppress the large combinatorial background from the two - jet like @xmath119 ( @xmath120 , @xmath121 , @xmath122 or @xmath123 ) continuum processes , variables that characterize the event topology are used .
we construct a fisher discriminant @xmath124 , from 6 modified fox wolfram moments @xcite .
furthermore , @xmath125 , the angle of the @xmath12 flight direction with respect to the beam axis is also used to distinguish signal from continuum background .
we combine these two independent variables , @xmath124 and @xmath125 to make a single likelihood ratio variable ( @xmath126 ) that distinguishes signal from continuum background .
we apply a different requirement for each sub - mode based on the expected signal yield and the backgrounds in the @xmath118 sideband data . for @xmath110 where @xmath127 , @xmath128 we require @xmath129 whereas for @xmath130 , @xmath131 , @xmath55 , @xmath56 , @xmath57 and @xmath58 we require @xmath132 . to give an example of the performance of this selection , the @xmath129 requirement keeps 87.5 % of the @xmath133\pi^{-}$]signal while removing 73 % of the continuum background .
the signal yields are extracted from a fit to the @xmath134 distribution in the region @xmath135 .
the @xmath110 signal is parameterized as a double gaussian with peak position and width floated . on the other hand , we calibrate the shape parameters of the @xmath112 signal using the @xmath36 data .
this accounts for the kinematical shifts and smearing of the @xmath134 peaks caused by the incorrect mass assignments of prompt hadrons .
the peak position and width of the @xmath136 signal events are determined by fitting the @xmath110 distribution using the kaon mass hypothesis for the prompt pion , where the relative peak position is reversed with respect to the origin .
the shape parameters for the feed - across from @xmath110 are fixed by the fit results of the @xmath110 enriched sample .
the continuum background is modeled as a first order polynomial function with parameters determined from the @xmath134 distribution for the events in the sideband region @xmath137 .
backgrounds from other @xmath12 decays including contributions from @xmath138 and @xmath139 are modeled as a smoothed histogram from monte carlo simulation .
the fit results are shown in fig . 2 . [ cols="<,^,^,^,^ " , ] |
the point form is one of the three prominent forms proposed by dirac in his seminal paper of 1949 @xcite to formulate relativistic hamiltonian dynamics .
it has the nice feature that the whole lorentz group ( rotations and boosts ) is kinematical , i.e. is not affected by interactions .
this allows to boost and rotate bound - state wave functions in a simple way . as a price ,
all components of the 4-momentum operator become interaction dependent .
the formalism presented here is based on the point form of relativistic quantum mechanics and makes use of the bakamjian - thomas construction @xcite for introducing interactions in a fully poincar invariant manner . as a consequence
the 4-momentum operator factorizes into an interacting mass operator and a free velocity operator so that it suffices to consider only an eigenvalue problem for the mass operator .
the formalism presented here has been applied successfully to the study of electromagnetic properties of spin-0 and spin-1 two - body bound states consisting of equal - mass particles @xcite , as well as to the electroweak structure of mesons consisting of constituents with different masses @xcite .
the starting point of all these calculations is the physical processes in which the form factors can be measured , i.e. electromagnetic scattering or weak decays . to describe these processes in a fully poincar invariant manner a multichannel version of the bakamjian - thomas construction @xcite
is employed . as mentioned already
, the 4-momentum operator then factorizes into an interacting mass operator and a free 4-velocity operator : @xmath2 the ( free ) 4-velocity operator @xmath3 is defined by @xmath4 and describes the overall motion of the system .
the mass operator @xmath5 , depending on internal variables only , is the quantity of interest , since it contains the information on the internal structure of the system .
electron - meson scattering , e.g. , is then formulated on a hilbert space consisting of a @xmath6 and a @xmath7 sector . a convenient basis consists of , so - called ,
_ velocity states_. these are multiparticle states characterized by the overall 4-velocity and the center - of - mass momenta and spins of its components @xcite .
the mass eigenvalue equation to be solved has the form : @xmath8 the diagonal elements of the mass matrix contain the relativistic kinetic energies and an instantaneous confining interaction between the quarks .
the transition between both channels is caused by @xmath9 and @xmath10 , which are vertex operators that account for the creation and annihilation of one photon , respectively .
they are uniquely related to the interaction lagrangian density of qed @xcite . eliminating the @xmath7 channel
one ends up with an equation for the @xmath6 component : @xmath11 @xmath12 is an optical potential that describes the ( dynamical ) 1-photon exchange between electron and ( anti)quark . on - shell matrix elements of @xmath12 between ( velocity ) states of a confined @xmath13 pair with quantum numbers of the meson @xmath14 provide the invariant 1-photon - exchange amplitude from which the electromagnetic current of the meson @xmath14 can be extracted : @xmath15 this relation determines the hadron current and thus the electromagnetic form factors in a unique way and it fixes also the normalization of the form factors .
it can be shown @xcite that the resulting electromagnetic current transforms covariantly under lorentz transformations and it is conserved for pseudoscalar mesons . however , because of cluster separability problems inherent in the bakamjian - thomas construction @xcite one observes that the current obtained in this way can not be decomposed in terms if hadronic covariants only , but one needs additional covariants built with the sum of the incoming and outgoing electron 4-momenta @xcite .
the need of such additional , spurious covariants resembles the situation in the covariant front - form formalism @xcite .
there one also encounters spurious dependencies of the currents on a 4-vector that specifies the orientation of the light front .
remarkably , if we let the invariant mass of the electron - meson system go to infinity our electromagnetic form factors turn out to agree with the front - form results computed in the @xmath16 frame @xcite .
it is quite obvious , how this formalism can be generalized to semileptonic weak decays in order to calculate transition amplitudes , currents and decay form factors @xcite .
decay processes involve time - like momentum transfers .
unlike scattering the covariant decomposition of decay currents does not require the introduction of spurious covariants , neither in pseudoscalar - to - pseudoscalar nor in pseudoscalar - to - vector meson decays .
form factors can be extracted unambiguously in a frame - independent way @xcite .
analytical and numerical studies of electromagnetic and weak form factors of heavy - light systems show that the predictions of heavy - quark symmetry are respected when one of the constituent masses goes to infinity @xcite .
as it should be , one ends up with the , so - called , isgur - wise function @xcite , i.e. one single , universal , spin - independent form factor that does not depend on the mass of the heavy quark . within our approach
the isgur - wise function acquires a simple analytical form : @xmath17 with @xmath18 and @xmath19 denoting the initial and final meson 4-velocities , respectively .
@xmath20 is a spin - rotation factor and @xmath21 with @xmath22 .
is by no means negligible , which emphasizes the need of an appropriate relativistic treatment of the spin rotation when boosting bound states . ]
as an example we present numerical results for the transition form factors of the semileptonic @xmath23 decay .
calculations are done with a simple harmonic - oscillator wave function ( see ref .
@xcite ) with oscillator parameteter @xmath24 gev .
the most general covariant decomposition of the weak @xmath25 transition current is usually written in the form : @xmath26 where @xmath27 .
as mentioned above , our microscopic expression for the @xmath28 transition current is not plagued by spurious contributions and the covarinat decomposition ( [ eq : jppsvdec ] ) can be applied directly to extract the decay form factors @xcite . introducing the shorthand notation @xmath29 and adopting the same kinematics as in ref .
@xcite one encounters 10 non - vanishing spin matrix elements of the current , namely @xmath30 , @xmath31 , and @xmath32 , @xmath33 . taking into account that @xmath34 and @xmath35 are related by parity , one
is left with only 6 different matrix elements , 4 of them being independent .
one can see that @xmath36 and @xmath37 enter only @xmath38 and @xmath39 .
thus the set @xmath30 , @xmath31 , @xmath38 and @xmath39 can be used to extract all the @xmath40 decay form factors .
the numerical results for the form factors as a function of @xmath41 ( multiplied by appropriate kinematical factors @xcite ) are shown and compared with the isgur - wise function and with the available experimental data in fig .
[ fig : decpsv ] .
our numerical values for the isgur - wise function agree with those obtained with a front - form quark model @xcite .
discrepancies between the point- and front - form approach show up as soon as the decay form factors are calculated for finite , physical masses of the heavy quarks .
most likely , these differences can be attributed to the different roles played by z - graphs , i.e. non - valence contributions , in either approach . in the heavy - quark limit z - graphs
do not contribute to the current , neither in the front form nor in the point form , which explains why the results agree for the isgur - wise function .
for finite quark masses , however , the inclusion of z - graphs seems to be crucial for the frame independence of the decay form factors in front form @xcite , whereas this is not the case in point form ( as discussed above ) .
decay form factors ( multiplied by appropriate kinematical factors , @xmath42 ) in comparison with the isgur - wise function @xmath43 and with the available experimental data .
the values taken for the physical quark masses are @xmath44 gev , @xmath45 gev and @xmath46 gev .
experimental data are taken from belle @xcite ( dots ) , cleo @xcite ( triangles ) and babar @xcite ( crosses ) assuming that @xmath47 , i.e. the central value given by the particle data group @xcite .
, title="fig:",scaledwidth=80.0% ] +
a poincar invariant description of electromagnetic and weak currents and form factors of two - body systems has been presented .
the predictions of heavy - quark symmetry are respected by this approach and a simple analytical expression for the isgur - wise function in terms of the initial and final wave function has been derived . as an example , numerical results for the semileptonic @xmath48 decay have been given .
direct comparison of the isgur - wise function with the transition form factors that are obtained for physical masses of the heavy quarks revealed that heavy - quark symmetry is broken by about @xmath49 .
it remains to be seen , whether z - graph contributions to the decay process may restore the equivalence of our point - form approach with front - form calculations that has already been found for electromagnetic form factors in the space - like momentum - transfer region @xcite .
m. gomez - rocha and w. schweiger , few body syst .
* 50 * , 227 ( 2011 ) [ arxiv:1011.0547 [ hep - ph ] ] .
m. gomez - rocha , e. .p .
biernat and w. schweiger , few body syst .
* 52 * , 397 ( 2012 ) [ arxiv:1110.2355 [ hep - ph ] ] .
w. h. klink , phys .
c * 58 * , 3617 ( 1998 ) .
j. carbonell , b. desplanques , v. a. karmanov and j. f. mathiot , phys .
* 300 * , 215 ( 1998 ) [ nucl - th/9804029 ] .
n. isgur and m. b. wise , phys .
b * 232 * , 113 ( 1989 ) .
n. isgur and m. b. wise , phys .
b * 237 * , 527 ( 1990 ) .
h. -y .
cheng , c. -y .
cheung and c. -w .
hwang , phys .
d * 55 * , 1559 ( 1997 ) [ hep - ph/9607332 ] .
b. l. g. bakker , h. -m .
choi and c. -r .
ji , phys .
d * 67 * ( 2003 ) 113007 [ hep - ph/0303002 ] .
k. abe _ et al . _ [ belle collaboration ] , phys .
b * 526 * ( 2002 ) 247 [ hep - ex/0111060 ] .
n. e. adam _ et al .
_ [ cleo collaboration ] , phys . rev .
d * 67 * ( 2003 ) 032001 [ hep - ex/0210040 ] .
b. aubert _ et al . _ [ babar collaboration ] , phys .
d * 71 * ( 2005 ) 051502 [ hep - ex/0408027 ] .
j. beringer _
( particle data group ) , phys .
d * 86 * ( 2012 ) 010001 . | we present a relativistic point - form approach for the calculation of electroweak form factors of few - body bound states . as an example , the transition form factors for the semileptonic weak decay @xmath0 are discussed and it is sketched how they can be extracted unambiguously from the invariant transition amplitude that describes the process .
it is shown how these form factors go over into one universal function , the isgur - wise function in the heavy - quark limit , @xmath1 , and comparison with the available experimental data is made . |
the bosonic d - brane has an unstable tachyonic mode on the perturbative vacuum .
it has been conjectured @xcite that there is a local minimum of the tachyon potential ( non - perturbative vacuum ) , where the tension of the d - brane and the negative contribution from the tachyon potential exactly cancel each other . on this vacuum , there should be no d - branes and hence no open string excitations at all , leaving only the closed string sector .
it has been also conjectured that the solutions on this non - perturbative vacuum represent various lower dimensional d - branes ( descent relations ) . for studying these conjectures , string field theories
have proved to be very useful , and there have appeared many interesting works in this direction @xcite@xcite . in particular , witten s cubic open string field theory ( csft ) @xcite combined with the level truncation approximation @xcite has been successful in verifying the conjecture on the potential height to a very high degree of accuracy @xcite . though most of these conjectures , the potential height problem and the descent relation , have later been proved @xcite in another formulation of open string field theory called boundary string field theory ( bsft ) @xcite , there still remains an important problem to be understood : how the open string excitations disappear around the non - perturbative vacuum .
note that for analyzing this problem we have to treat all the modes in open string theory other than the tachyon mode .
therefore , csft whose action is given in a closed form seems to be rather suited for this problem than bsft which has concise expression only for the tachyon field .
one of possible ways to prove the absence of physical open string excitations is to show that they all have infinite masses @xcite .
another and roughly an equivalent way is to show that the physical components of the open string field fluctuations lack their kinetic terms in the sft action and hence become non - dynamical @xcite ( kinetic term means the quadratic term of the fluctuations containing the derivatives with respect to the center - of - mass string coordinate ) . in relation to this , an interesting deformed version of csft has recently been proposed in @xcite as a possible sft model around the non - perturbative vacuum .
the brst operator in this model is simply equal to the ghost zero - mode , @xmath0 , and hence the quadratic term @xmath1 contains no kinetic term at all .
what we have to do for proving the absence of open string excitations on the non - perturbative vacuum is to derive this kind of sft from the ordinary csft .
the purpose of this paper is to explore , using the level truncation approximation , the possibility that csft action expanded around the non - perturbative vacuum contains no kinetic terms for physical fluctuations .
let the action of csft be given by @xmath2 , and let us expand the original string field @xmath3 as @xmath4 with @xmath5 being the classical solution representing the non - perturbative vacuum .
then the term in @xmath6 quadratic in the fluctuation @xmath7 is @xmath8 .
since the three - string interaction vertex defining the star - product @xmath9 contains any powers of the derivative @xmath10 with respect to the center - of - mass @xmath11 of the string coordinate @xmath12 , the kinetic term for the fluctuation @xmath7 consists of terms with any number of derivatives .
our expectation is that these kinetic terms all vanish for the physical components of @xmath7 , where `` physical '' means being brst invariant .
we are going to study this possibility in the level ( 2,6 ) truncation by taking as the fluctuations only the scalar modes at levels zero and two . as the kinetic terms we take the quadratic and the quartic ones in the derivatives @xmath10 .
as the simplest and encouraging example , let us quote our result in the level @xmath13 approximation , namely , let us take only the level zero scalar field @xmath14 ( see sec .
the kinetic term for @xmath14 quadratic in the derivative is @xmath15 \left({\partial}_\mu t\right)^2 , \label{eq : skin}\ ] ] where @xmath16 is the classical value of @xmath14 at the non - perturbative vacuum .
this kinetic term vanishes at @xmath17 , which is close to the value of @xmath16 obtained as the stationary point of the level ( 0,0 ) potential , @xmath18 .
even this simplest example supports our expectation that the kinetic term for the physical fluctuations totally vanish . in our analysis in the level ( 2,6 ) approximation
, we consider the kinetic terms quadratic and quartic in the derivatives for the three scalar fluctuation modes at levels zero and two .
then the coefficients of the kinetic terms form @xmath19 matrices in the space of fluctuations .
since there is only one physical component among the three scalar fluctuations , the two matrices , one for the kinetic term with two derivatives and the other with four , must have a common zero - mode ( namely , the eigenvector corresponding to the eigenvalue zero ) in order for the physical fluctuation to disappear from the kinetic terms .
we examine numerically whether this condition is satisfied in the neighborhood of the non - perturbative vacuum , and obtain results which support our expectation .
the organization of the rest of this paper is as follows . in sec . 2 , after making a setup for the level truncation calculation in csft , we present the derivative expansion of the kinetic term of the action and the brst transformation at a generic translation invariant background .
then , in sec .
3 we carry out the numerical analysis of the kinetic terms and the brst transformations of the fluctuations to examine whether the kinetic terms for the physical fluctuation can disappear in the neighborhood of the non - perturbative vacuum . finally in sec . 4
, we discuss future problems .
in this section , as a preparation for the analysis in sec . 3 , we fix our conventions for the csft action , and then present the derivative expansion of the kinetic terms and the brst transformation for the scalar fluctuation modes in the level ( 2,6 ) approximation .
we shall first recapitulate the gauge - fixed action and the brst transformation of csft in the siegel gauge .
they are both obtained from the gauge invariant action : @xmath20 where @xmath21 is the open string coupling constant , and @xmath22 denotes the integrations over the anti - ghost zero - mode @xmath23 and the center - of - mass coordinate @xmath11 of the @xmath24-th string .
the string field @xmath25 in ( [ eq : sinv ] ) is a state in the fock space of the first quantized string and carry ghost number @xmath26 .
it is expanded in terms of the anti - ghost zero mode @xmath23 as @xmath27 where @xmath28 and @xmath29 carry ghost number @xmath30 and @xmath26 , respectively .
the ghost zero - mode structure of the brst operator @xmath31 is @xmath32 where we have @xmath33 with @xmath34 .
the three - string interaction vertex @xmath35 is given in the momentum representation for @xmath11 as @xmath36{|\ , 0\ , \rangle_{123 } } { \nonumber}\\ & \hspace*{5cm}\times ( 2\pi)^{26}\delta^{26}\left(p_1+p_2+p_3\right ) , \end{aligned}\ ] ] where @xmath37 is the fock vacuum satisfying @xmath38 ( @xmath39 ) , @xmath40 is a constant @xmath41 @xcite , and the neumann coefficients , @xmath42 and @xmath43 , are defined in @xcite . in particular , for @xmath44 we have @xmath45 with @xmath46 .
then the gauge - fixed action @xmath6 in the siegel gauge , @xmath47 , is obtained from @xmath48 ( [ eq : sinv ] ) simply by putting @xmath49 : @xmath50 where the three - string vertex for the gauge - fixed action is @xmath51 .
this action has an invariance under the brst transformation @xmath52 : @xmath53 physical quantities in the present sft are required to be invariant under the brst transformation . as mentioned in sec . 1
, we are interested in the kinetic term of the string field fluctuation around the non - perturbative vacuum . to analyze it in the simplest framework
, we adopt the level ( 2,6 ) truncation and take the scalar fields at levels zero and two , both as the vacuum coordinates and the fluctuations .
namely , we restrict the @xmath54 component field sector of @xmath28 to the following form : @xmath55 then , reexpressing the three component field @xmath56 as @xmath57 with @xmath58 being a ( translation invariant ) vacuum value and @xmath59 the fluctuation around @xmath58 , we carry out the derivative expansion of the kinetic term of the fluctuation @xmath59 obtained from the action @xmath6 ( [ eq : s ] ) : @xmath60 here we are making the simplification of considering only the scalar fields @xmath61 . in particular , we took only the trace part of the tensor fields @xmath62 which appear in the expansion of @xmath28 in the form @xmath63 .
we have also discarded the vector field at level two which appear as the coefficient of @xmath64 .
complete treatment in the ( 2,6 ) truncation should take into account the mixing among all the scalar , vector and tensor fields at levels zero , one and two
. however , such analysis would be so complicated since the size of the matrices @xmath65 and @xmath66 become much larger , and there are many physical fields that should be confined . therefore , we shall carry out the following analysis in the smallest fluctuation space @xmath61 with a hope that the mixing with other components is small .
now each component of the @xmath19 matrix @xmath67 is a linear function of @xmath68 and is given explicitly by ( hereafter we omit the subscript 0 on @xmath68 unless confusion occurs ) @xmath69 note that in the present convention we have @xmath70 at the perturbative vacuum ( the minus sign for the @xmath71 field is due to the negative norm property of the ghosty state @xmath72 ) .
the matrix @xmath73 for the four - derivative kinetic term have contributions only from the interaction term of ( [ eq : s ] ) .
it is given by @xmath74 ^ 2}\left ( 27\,t + 11\,u - 5\sqrt{13}\,v\right ) , \\ h_{uu}=&-\frac{19}{54\sqrt{39}}{\ln\!\left(\frac{3 ^
3}{2 ^ 4}\right)}v + \frac{\sqrt{3}}{64}{\left[\ln\!\left(\frac{3 ^ 3}{2 ^ 4}\right)\right]^2}\left ( \frac{19}{3}\,t + u - \frac{95\sqrt{13}}{81}\,v\right ) , \\ h_{vv}=&\frac{1}{\sqrt{3}}\left(\frac{1}{78}\,t + \frac{11}{2106}\,u -\frac{5}{18\sqrt{13}}\,v\right ) + \frac{1}{156\sqrt{3}}{\ln\!\left(\frac{3 ^ 3}{2 ^ 4}\right)}\left ( 49\,t + \frac{539}{27}\,u -\frac{2779}{3\sqrt{13}}\,v\right ) { \nonumber}\\ & + \frac{1}{192\sqrt{3}}{\left[\ln\!\left(\frac{3 ^ 3}{2 ^ 4}\right)\right]^2}\left ( 581\,t + \frac{6391}{27}\,u -\frac{20951}{9\sqrt{13}}\,v\right ) , \\
h_{tu}= & -\frac{11}{6\sqrt{39}}{\ln\!\left(\frac{3 ^ 3}{2 ^ 4}\right)}v + \frac{\sqrt{3}}{64}{\left[\ln\!\left(\frac{3 ^ 3}{2 ^ 4}\right)\right]^2}\left ( 33\,t + \frac{19}{3}\,u -\frac{55\sqrt{13}}{9}\,v\right ) , \\
\frac{22}{1053\sqrt{3}}\,v + \frac{1}{24\sqrt{3}}{\ln\!\left(\frac{3 ^ 3}{2 ^ 4}\right)}\left(-\frac{11}{\sqrt{13}}\,t -\frac{19}{9\sqrt{13}}\,u + \frac{4279}{351}\,v\right ) { \nonumber}\\ & + \frac{1}{192\sqrt{3}}{\left[\ln\!\left(\frac{3 ^ 3}{2 ^ 4}\right)\right]^2}\left ( -55\sqrt{13}\,t - \frac{95\sqrt{13}}{9}\,u + \frac{6391}{27}\,v \right ) , \\ h_{vt}= & \frac{2}{39\sqrt{3}}\,v + \frac{\sqrt{3}}{8}{\ln\!\left(\frac{3 ^ 3}{2 ^ 4}\right)}\left(-\frac{3}{\sqrt{13}}\,t -\frac{11}{9\sqrt{13}}\,u + \frac{389}{117}\,v\right ) { \nonumber}\\ & + \frac{\sqrt{3}}{64}{\left[\ln\!\left(\frac{3 ^ 3}{2 ^ 4}\right)\right]^2}\left(-15\sqrt{13}\,t -\frac{55\sqrt{13}}{9}\,u + \frac{581}{9}\,v\right ) .\end{aligned}\ ] ] for completeness we present the potential @xmath75 in the ( 2,6 ) truncation : @xmath76 in the analysis of sec . 3
, we also need the derivative expansion of the brst transformation of the fields @xmath61 for identifying their physical ( brst invariant ) combination at the non - perturbative vacuum .
since the ghost field in @xmath28 at level one does not contribute to @xmath77 , we have only to consider the ghost fields at level two , @xmath78 and @xmath79 , which appear in @xmath28 as @xmath80 then , @xmath77 can be calculated using the formula ( [ eq : delb ] ) . similarly to the kinetic term ( [ eq : expands ] ) , we carry out the derivative expansion for the contribution from the last term of ( [ eq : delb ] ) and express the result as @xmath81 where we have omitted those terms containing the fluctuation @xmath59 on the rhs .
the explicit expression of the coefficients @xmath82 are as follows : @xmath83 note that the constant terms in @xmath84 , @xmath85 , @xmath86 and @xmath87 came from @xmath88 in ( [ eq : delb ] ) , and that the tachyon field @xmath14 is brst invariant at the perturbative vacuum .
in this section , we shall study numerically whether the conditions for the absence of the kinetic terms of the physical scalar field are realized near the non - perturbative vacuum .
the precise expressions of the conditions will be stated below , in particular , in sec .
. what we shall test are i ) location of the zero of @xmath89 and that of @xmath90 , ii ) coincidence of the zero - modes of @xmath65 and @xmath66 , and iii ) existence of the brst invariant linear combination of @xmath61 including the higher derivative terms in @xmath91 .
unfortunately , these conditions are not satisfied to high precision at the non - perturbative vacuum determined as a stationary point of the potential .
therefore , we shall look at each of them globally in the space of vacuum coordinate @xmath92 to test whether there is a point satisfying the condition near the non - perturbative vacuum .
let us first consider the simplest ( 0,0 ) truncation , namely , let us keep only the level zero field @xmath14 . in this approximation
the field @xmath14 is physical everywhere , @xmath93 , and the potential @xmath94 ( [ eq:26v ] ) and the metric @xmath95 ( [ eq : gtt ] ) are reduced to @xmath96 the non - trivial stationary point ( non - perturbative vacuum ) of @xmath97 exists at @xmath98 , while the metric @xmath99 vanishes at @xmath100 .
surprisingly these two points are very close , supporting that the kinetic term of the physical field @xmath14 is missing . as for @xmath101 for the four - derivative kinetic term
, it vanishes only at @xmath102 .
however , the level ( 0,0 ) approximation would be too poor for such higher derivative kinetic terms . encouraged by the above result in the ( 0,0 ) approximation , let us proceed to the analysis of the full @xmath65 of eqs .
( [ eq : gtt ] ) ( [ eq : gvt ] ) and the first two terms of the brst transformation ( [ eq : expanddelb ] ) : @xmath104 namely , we shall examine the lowest derivative kinetic term and the lowest derivative part of the brst transformation . before starting the analysis , let us state our expectation for the disappearance of the physical open string modes in the present situation .
since we have two independent ghost fields , @xmath105 and @xmath106 , on the rhs of ( [ eq : expanddelbab ] ) , there is only one brst invariant linear combination @xmath107 of the three scalar fields @xmath61 satisfying @xmath108 : @xmath109 where @xmath110 ( @xmath111 ) is the solution of @xmath112 namely , the level zero scalar @xmath14 , which was brst invariant at the perturbative vacuum , needs a mixing of @xmath71 and @xmath113 to reconstruct a brst invariant field at the non - perturbative vacuum .
a possible mechanism that confines this physical scalar @xmath107 is that @xmath107 has no kinetic term . in particular
, @xmath107 should not appear in @xmath114 .
for this to be realized , it is sufficient that the matrix @xmath115 has a zero - mode @xmath116 ( i.e. , the eigenvector of the eigenvalue zero ) and that it has a non - vanishing inner product with @xmath117 : @xmath118 the reason is as follows
. let @xmath119 ( @xmath120 ) be the three orthonormal eigenvectors of the ( real and symmetric ) matrix @xmath65 corresponding to the eigenvalue @xmath121 : @xmath122 then , the two - derivative kinetic term is diagonalized as @xmath123 where the new field @xmath124 is defined by @xmath125 .
in order for the kinetic term ( [ eq : kineticpsi ] ) not to contain the physical @xmath126 , @xmath107 should not be expressed as a linear combination of @xmath124 ( @xmath127 , and this is realized if the condition ( [ eq : betae0 ] ) holds .
we would like to emphasize that the missing combination @xmath128 need not be exactly equal to @xmath107 . is not invariant under a transformation of the coordinate system @xmath59 of the field space . ]
a rather weak condition ( [ eq : betae0 ] ) is sufficient .
therefore , what we have to test first of all is whether the determinant of the matrix @xmath115 vanishes near the non - perturbative vacuum .
recall that the coordinate @xmath129 of the non - perturbative vacuumobtained as a stationary point of the level ( 2,6 ) potential ( [ eq:26v ] ) is @xcite @xmath130 first , the value of @xmath89 at the point of ( [ eq : varphi_c ] ) is @xmath131 , which is relatively close to zero compared with its value at the perturbative vacuum @xmath132 , indicating that the point ( [ eq : varphi_c ] ) is near the two - dimensional surface of @xmath133 , though not exactly on it .
( 317,148)@xmath134 ( 20,258)@xmath135 ( 290,253)(@xmath136 ) ( 310,238)(@xmath137 ) ( 310,225)(@xmath138 ) ( 290,200)@xmath139 figs .
[ fig : plotdetg ] and [ fig : detgzerocurve ] would help understanding how the non - perturbative vacuum is close to the surface @xmath133 . in fig . [
fig : plotdetg ] we plot @xmath89 as a function of @xmath14 for various fixed values of @xmath140 .
in particular , the zero of @xmath89 for @xmath141 of ( [ eq : varphi_c ] ) is at @xmath142 , while the zero for @xmath143 is at @xmath144 .
, @xmath89 has other two zeros at @xmath145 and @xmath146 .
when @xmath143 , the other two zeros are complex . ] on the other hand , fig .
[ fig : detgzerocurve ] shows curves of @xmath133 in the @xmath140 plane for several fixed values of @xmath14 . as a concrete example , assuming the conjectured equality @xmath147 at the non - perturbative vacuum @xcite , the value of @xmath71 satisfying @xmath148 is @xmath149 .
( 290,120)@xmath150 ( 21,230)@xmath151 ( 285,220)@xmath152 ( 285,190)@xmath153 ( 285,163)@xmath154 ( 285,140)@xmath155 ( 285,105)@xmath156 as a representative point on the @xmath133 surface which is reasonably close to the non - perturbative vacuum of ( [ eq : varphi_c ] ) , there is a stationary point @xmath157 of the level ( 2,6 ) potential ( [ eq:26v ] ) restricted on the surface @xmath133 : @xmath158 the value of the potential ( [ eq:26v ] ) ( normalized so that the true value is equal to @xmath26 ) at this point is @xmath159 , while that at the point ( [ eq : varphi_c ] ) is @xmath160 .
the above analysis tells that the non - perturbative vacuum is indeed close to the surface @xmath133 .
in particular , for @xmath161 the condition @xmath133 combined with the assumption @xmath162 predicts the expected value @xmath163 for @xmath71 and @xmath113 . we have also examined the condition ( [ eq : betae0 ] ) to find that the inner product @xmath164 on the surface @xmath133 has no zeros at least in the region @xmath165 .
next let us consider the four - derivative kinetic term in ( [ eq : expands ] ) .
what we expect for @xmath73 in order for the physical scalar @xmath107 ( [ eq : t ] ) to drop out from the four - derivative kinetic term is as follows : * @xmath66 has a zero - mode @xmath167 at the non - perturbative vacuum .
* the two non - zero - modes of @xmath66 span the same two - dimensional subspace as that spanned by the non - zero - modes of @xmath65 ( equivalently , @xmath167 is parallel to the zero - mode @xmath116 of @xmath65 ) .
if these conditions are satisfied , then @xmath168 is also expressed only in terms of @xmath124 ( @xmath169 ) appearing in the two - derivative kinetic term ( [ eq : kineticpsi ] ) .
first , let us see whether @xmath90 vanishes near the non - perturbative vacuum . fig .
[ fig : plotdeth ] shows @xmath90 as a function of @xmath14 for several fixed values of @xmath140 . for @xmath143
, @xmath90 has no zeros in the range of @xmath14 of our interest , and for @xmath170 of ( [ eq : varphi_r ] ) , @xmath90 become close to zero at @xmath171 but does not cross zero . for larger values of @xmath113 compared with @xmath71 , @xmath172 and @xmath173 , @xmath90 has zeros around @xmath174 .
( 317,175)@xmath134 ( 0,258)@xmath175 ( 106,55)(@xmath136 ) ( 157,75)(@xmath176 ) ( 210,55)(@xmath177 ) ( 275,75)(@xmath178 ) ( 290,140)@xmath151 ( 18,232)@xmath150 ( 275,90)@xmath179 ( 265,58)@xmath174 ( 215,50)@xmath180 fig .
[ fig : resolhu ] shows the curves of @xmath181 in the @xmath182 plane for three values of @xmath14 . from fig .
[ fig : resolhu ] , we see that , for @xmath161 and @xmath183 corresponding to the non - perturbative vacuum , the @xmath113 coordinate on the @xmath181 surface is @xmath184 , which is more than 50% larger than the value of the non - perturbative vacuum . in spite of this discrepancy , it seems miraculous that the surface @xmath181 is rather close to the non - perturbative vacuum .
therefore , expecting that the discrepancy would be resolved in a more precise analysis , let us proceed to the study of another condition , @xmath185 .
ideally , we should test whether the two vectors @xmath116 and @xmath167 become parallel at some point on the curve of the intersections of the two surfaces @xmath133 and @xmath181 .
however , though the two surfaces are separately close to the non - perturbative vacuum , their intersection curve is not so close to it in the present approximation . in the range @xmath186 , the corresponding @xmath14 and @xmath113 coordinates on the intersection curve do not change rapidly and take values @xmath187 and @xmath188 . for large values of @xmath71 , the corresponding @xmath113 coordinate become outside the range @xmath189 .
] therefore , we instead carry out an indirect analysis .
as @xmath116 , we take the zero - mode of @xmath115 at a fixed @xmath190 which is on the surface @xmath133 and close to the non - perturbative vacuum . concretely , we take @xmath191 of ( [ eq : varphi_r ] ) , and the corresponding zero - mode is @xmath192 .
then , we plot , on the surface @xmath181 , the absolute value of the inner products @xmath193 and @xmath194 , where @xmath195 and @xmath196 are the normalized eigenvectors of @xmath66 corresponding to non - zero eigenvalues . on the intersection curve ,
the condition @xmath185 is equivalent to @xmath197 .
we are going to examine the behavior of the approximate quantities to the two inner products on the surface @xmath181 .
( 288,65)@xmath151 fig .
[ fig : fe ] shows these two inner products as a function of @xmath113 with @xmath14 fixed to @xmath198 and @xmath71 determined by @xmath181 .
note that one of the inner products vanishes at @xmath199 and the other at @xmath200 .
moreover , the positions of these two zeros , in particular , @xmath199 , are fairly insensitive to the choice of @xmath116 used for taking the inner products .
namely , the positions of the zeros change little even if we replace @xmath201 by almost any vector .
this is ascribed to the fact that the eigenvectors @xmath195 and @xmath196 change their directions rapidly by large angle in the neighborhood of @xmath199 and that of @xmath200 , respectively ( recall that @xmath202 vanishes when @xmath203 crosses the surface perpendicular to @xmath116 ) .
although the inner product @xmath204 is not invariant under the change of the field space coordinates @xmath59 , the positions of the zeros are insensitive to the coordinate change owing to the same property of @xmath205 .
for example , if we adopt the coordinates of @xcite , @xmath206 , the global behavior of the inner products differs from fig . [
fig : fe ] , but there are zeros near the corresponding points . therefore , the presence of the two zeros of the inner products @xmath207 , in particular , the zero at @xmath199 which is closer to the non - perturbative vacuum , is very encouraging .
it is expected that , as we improve the approximation , the two zeros approach to each other and to @xmath208 .
we have defined the physical scalar field @xmath107 ( [ eq : t ] ) by the requirement that it be invariant under the brst transformation @xmath209 ( [ eq : expanddelbab ] ) truncated at the lowest order in the derivatives on @xmath105 and @xmath106 . however , as given in ( [ eq : expanddelb ] ) , the full brst transformation contains higher derivative terms , and we have to check whether the @xmath210 and @xmath211 terms are also cancelled in @xmath212 .
namely , we shall test whether @xmath213 and @xmath214 , @xmath215 with @xmath216 satisfying ( [ eq : eqforbeta ] ) vanish near the non - perturbative vacuum .
( 288,155)@xmath150 ( 22,230)@xmath151 ( 288,130)@xmath134 ( 10,230)*@xmath217 , @xmath218 * ( 210,217)(@xmath138 ) ( 280,210)(@xmath219 ) ( 240,180)(@xmath220 ) let us first consider @xmath213 on the surface @xmath221 . fig .
[ fig : zerotesty ] shows the curve in the @xmath140 plane determined by @xmath222 and @xmath221 .
note that this curve passes the point @xmath223 , implying that @xmath222 holds near the non - perturbative vacuum . however , the corresponding curve for @xmath214 is totally outside the region of our interest in the @xmath140 plane .
therefore , we instead studied the behavior of @xmath214 as a function of @xmath14 for a number of fixed values of @xmath140 ( see fig . [
fig : testyzall ] ) . from fig .
[ fig : testyzall ] we see that @xmath214 has a zero at a rather small @xmath14 when the @xmath140 coordinate is close to that of the non - perturbative vacuum . in this section ,
we have analyzed the conditions for the absence of the kinetic term of the brst invariant scalar fluctuation .
the conditions can simply be stated as ( i ) @xmath65 and @xmath66 have a common zero - mode at the non - perturbative vacuum , and ( ii ) the brst invariant fluctuation component can be consistently defined there .
we divided the test of the conditions into many steps and examined whether the following quantities vanish near the non - perturbative vacuum : @xmath89 ( sec .
3.2 ) , @xmath90 ( sec .
3.3 ) , @xmath224 ( sec .
3.3 ) , @xmath225 and @xmath226 ( sec .
3.4 ) . among them
, we obtained fairly good results for @xmath89 , @xmath90 , @xmath227 ( the one having a zero at @xmath199 in fig .
[ fig : fe ] ) and @xmath225 . however , the behaviors of @xmath228 and @xmath226 were not as we expected .
they are hoped to improve in a better treatment , in particular , by taking into account the mixing with the vector and tensor components at level two which we neglected in this paper .
we have carried out the same analysis also in the level ( 2,4 ) approximation , and found that the results are not qualitatively changed .
our analysis in this paper is incomplete in various aspects , and further detailed studies are needed for confirming the absence of kinetic terms .
we have to extend the space of fluctuations to incorporate the vector and tensor components even in the level ( 2,6 ) approximation .
this is indispensable for making clear the full brst structure of the fluctuations .
we have to consider the kinetic terms with more than four derivatives .
and we have to raise the order of the level truncation .
what is more important is of course the analytical understanding of the absence of the kinetic terms , though this must be a difficult task since the exact solution @xmath5 for the non - perturbative vacuum would be needed .
conversely , the missing kinetic terms for the fluctuations could give a hint for finding the exact solution .
it is also an interesting question whether the csft expanded around the non - perturbative vacuum is a topological theory described by a brst exact action @xcite .
we have to clarify how such an open sft can support the closed string sector .
we would like to thank m. bando , i. kishimoto , t. kugo , s. moriyama , t. noguchi and s. shinohara for valuable discussions and useful comments .
we would also like to thank s. moriyama for careful reading of the manuscript . the work of h. h. is supported in part by grant - in - aid for scientific research from ministry of education , science , sports and culture of japan ( # 12640264 ) . | it has been conjectured that the bosonic open string theory around the non - perturbative tachyon vacuum has no open string dynamics at all .
we explore , in the cubic open string field theory with level truncation approximation , the possibility that this conjecture is realized by the absence of kinetic terms of the string field fluctuations .
we study the kinetic terms with two and four derivatives for the lower level scalar modes as well as their brst transformation properties .
the behavior of the coefficients of the kinetic terms in the neighborhood of the non - perturbative vacuum supports our expectation that the brst invariant scalar component lacks its kinetic term .
= 18.6pt plus 0.2pt minus 0.1pt addtoresetequationsection |
the field of non - equilibrium statistical physics is related to those problems for which a hamiltonian formulation is not possible or the detailed - balance condition is not satisfied .
one of these problems is the study of surface growth and interfaces , where the understanding of the basic mechanisms is the key to developing thin - film devices with important technological applications @xcite . in this work
we are interested in the description of the morphology of a surface formed by adding particles to an initially flat substrate .
particles , having one unit height and linear size @xmath5 , land horizontally onto the surface , or over a straigth line in the case of a one - dimensional substrate , and are not allowed to diffuse . in this way
, we calculate the interface width , @xmath6 , a function that determines the roughness of the interface , where @xmath7 is the linear size of the substrate and @xmath8 is the time variable .
the simplest known model for the deposition of particles is the random deposition , where particles are aggregated onto an initially flat substrate , and lateral correlations among the deposited particles are completely neglected . for this trivial model ,
its continuous and discrete atomistic versions have exact solutions .
the simplest non - trivial growth models , which incorporate spatial and temporal correlations , can be described by means of stochastic differential equations .
the edwards - wilkinson ( ew ) equation @xcite is designed to take into account the surface relaxation and height - height correlations in some deposition models . on the other hand , when we need to include some lateral growth of the interface , as in the ballistic deposition of particles @xcite , the appropriate stochastic equation is the non - linear kardar - parisi - zhang ( kpz ) equation @xcite .
the ( ew ) and ( kpz ) equations can be written as @xmath9 \label{ew}\ ] ] and @xmath10 , \label{kpz}\ ] ] where @xmath11 represents the random fluctuations in the deposition processes . in order to calculate the interface width , we determine the vertical height of a given point of the surface relative to the substrate , @xmath12 , where * r * denotes the position of a cell over the substrate .
the roughness @xmath6 is defined as the mean square fluctuation of the height , @xmath13^{2}\rangle^{1/2}$ ] , where @xmath14 is the average value of the surface height at a given instant of time @xmath8 .
it is well established that for a large class of growth models the family - vicsek scaling relation applies @xcite @xmath15 where the scaling function @xmath16 is a constant for very large values of @xmath17 , and @xmath18 when @xmath19 .
the exponent @xmath1 characterizes the interface width at long times , @xmath20 is the dynamic exponent , while @xmath21 is the exponent associated to the initial stages of the growth .
these exponents are not at all independent , and are related by @xmath22 . in the theoretical studies performed in the area of surface growth ,
one is interested in the calculation of these exponents .
there are many studies in the literature concerning the properties of the growth models where calculations have been done analytically , or through the extensive use of monte carlo simulations @xcite . in this work we use monte carlo simulations to study the surface growth in @xmath23 and @xmath24 dimensions due to the deposition of particles of different sizes .
there are in literature some studies where two or more different deposition models are combined @xcite , or two different species of particles are deposited @xcite in order to describe real systems . here
, we choose our particles to be deposited from a modified poisson distribution .
this type of distribution appears to be relevant in some ash particles deposition on the heat exchange surfaces @xcite .
recently , we have investigated the surface growth generated by the random deposition of particles of different sizes in one and two dimensions @xcite .
the results of our simulations have shown that in both dimensions , the roughness evolves in time following three different behaviors .
the roughness at the initial times behaves as in the random deposition model , where correlations are absent . at intermediate times
, the surface roughness depends on the dimensionality of the system , and finally , at long times , it enters into the saturation regime .
however , there is a remarkable difference in the results we find in one and two dimensions , regarding the behavior of the interface width at the intermediate times . while in one dimension , the growth exponent is independent of the particle size , in two dimensions it increases with the size of the particle .
this happens because when we try to deposit a new particle in one dimension , it can be accomodated in one of the two ends of an already deposited particle , independently of its size .
however , in two dimensions , the new deposition can be done in the @xmath25 positions around a deposited particle , where @xmath5 is the linear size of the particle .
this dependence on @xmath5 makes the growth in two dimensions to be non - universal .
in the present study particles of linear size @xmath5 are dropped randomly over a finite one- or two - dimensional substrate .
all the particles considered in the deposition process are one unit height and length sizes in the range @xmath26 .
the most important difference between this model and the simple random deposition model , is that our model naturally allows for correlations among the columns , that is , it leads to a lateral growth of the interface .
figure [ deposition ] is a graphic representation of the one - dimensional deposition of a mixture of particles , which are selected from a modified poisson distribution , where the maximum particle size is @xmath27 . .
in this figure particles are selected from a modified poisson distribution of mean value @xmath28 .
time grows from bottom to top , where the interface width reached the stationary state . ] in two dimensions , a particle is aggregated only if the chosen site for deposition coincides with the midpoint of the particle , and there is enough space to accomodate it in the @xmath17 or @xmath29 directions .
otherwise , if the cell is already occupied or there is no enough space for the deposition , the particle is reflected away due to these geometric constraints .
particles are not allowed to diffuse or share the same cell in any deposition plane .
when we are depositing identical particles , during a unit of time , we try to deposit @xmath30 particles , which roughly means the deposition of one layer . for the deposition of particles of same size @xmath5 in a one - dimensional substrate , during a unit of time we try to deposit @xmath31 particles .
therefore , depending on the size of the particle , many trials of deposition are not successful in the unit of time .
in this section we only show some plots for the deposition in two dimensions .
our monte carlo simulations were performed on square lattices with linear size ranging from @xmath32 to @xmath33 .
we exhibit in figure [ roughnessxt ] the log - log plot of the interface width as a function of time for particles of size @xmath34 . at the initial times of deposition
, the growth exponent @xmath35 is close to that of the random deposition . at intermediate times
, the growth exponent exhibits a completely different behavior from the one observed in one dimension @xcite . in one dimension , the growth exponent @xmath2 at intermediate times , does not depend on the size of particle and its value is @xmath36 . in two dimensions ,
this exponent exhibits a very peculiar behavior . while for @xmath37
its value is around @xmath38 , which is smaller than @xmath35 , it jumps , for @xmath39 , to the value @xmath40 .
for @xmath41 it increases , but the rate of increasing is lower for higher values of @xmath5 . ,
and the linear size of the square lattice is @xmath42 . ]
we show in figure [ roughness ] the behavior of the interface width for the deposition on a square substrate of linear size @xmath42 for a mixture of particles with sizes in the range @xmath26 . every time we try to deposit a new particle , it is selected from a modified discrete poisson distribution whose maximum value is @xmath27 . as to be expected the figures we find for the scaling exponents are intermediary between those calculated for the smallest and largest particles . on a square lattice of side @xmath42 .
the values of the growth exponents @xmath35 and @xmath2 are indicated in the plots . ] in figure 4 we display the values of the growth exponent @xmath2 as a function of @xmath5 for three different lattice sizes .
we see that it is almost independent of @xmath7 , but the rate of increasing is very small for higher values of @xmath5 .
although not shown , we have also calculated the roughness exponent @xmath1 by extrapolating the saturation value of the interface width observed at long times for large values of @xmath7 and different particles sizes .
we have found that @xmath1 is very close to @xmath43 .
for the deposition on a linear substrate we have determined the value @xmath44 , which is also independent of @xmath5 .
therefore , in one dimension , using this value of @xmath1 , along with @xmath36 , the family - vicsek relation gives directly the value @xmath45 , for any value of particle length @xmath5 .
these figures put the one - dimensional version of this model in the same universality class of the villain - lai - das sarma equation @xcite . despite @xmath1
is close to @xmath43 in two dimensions , the family - vicsek relation predicts values for the dynamic exponent @xmath20 decreasing with the size @xmath5 of the particles .
then , the behavior of the model in two dimensions is nonuniversal . as a function of the particle length @xmath5 for three different linear sizes @xmath7 of the square lattice as indicated in the figure.,title="fig : " ] +
we employed monte carlo simulations to study the surface growth due to the deposition of rod - like particles of different sizes over linear and square lattices.we have shown that the interface width as a function of time exhibits three different behaviors . at the initial times
its behavior is similar to the usual random deposition model . at intermediate times
, the surface roughness depends on the system dimensionality and finally , at long times , it saturates .
we have seen that in one dimension the dynamics of the model can be mimicked by the villain - lai - das sarma equation for whatever particle size . on the other hand , for the deposition in two dimensions ,
the growth exponent at intermediary times changes with the particle size and the deposition model behaves differently from one dimension . for the deposition in two dimensions , we show that the interface width in the second regime presents an unusual behavior , described by a growth exponent @xmath2 , which depends on the size of the particles added to the substrate .
if the linear size of the particle is two , we found that @xmath3 , otherwise it is @xmath4 , for all particles sizes larger than three . while in one dimension the scaling exponents are the same as those predicted by the villain - lai - das sarma equation , in two dimensions , the growth exponents are nonuniversal . | monte carlo simulations are employed to investigate the surface growth generated by deposition of particles of different sizes on a substrate , in one and two dimensions .
the particles have a linear form , and occupy an integer number of cells of the lattice .
the results of our simulations have shown that the roughness evolves in time following three different behaviors .
the roughness in the initial times behaves as in the random deposition model , with an exponent @xmath0 . at intermediate times , the surface roughness depends on the system dimensionality and , finally , at long times , it enters into the saturation regime , which is described by the roughness exponent @xmath1 .
the scaling exponents of the model are the same as those predicted by the villain - lai - das sarma equation for deposition in one dimension . for the deposition in two dimensions ,
we show that the interface width in the second regime presents an unusual behavior , described by a growing exponent @xmath2 , which depends on the size of the particles added to the substrate .
if the linear size of the particle is two , we found that @xmath3 , otherwise it is @xmath4 , for all particles sizes larger than three . while in one dimension the scaling exponents are the same as those predicted by the villain - lai - das sarma equation , in two dimensions , the growth exponents are nonuniversal .
_ keywords : growth models , surface roughness , film deposition , random deposition , computer simulations . _ |
recently , the gravitational waves which are the solutions of the einstein equations and are the part of the most interesting astrophysical events were observed by ligo @xcite .
the einstein equations are sets of the nonlinear partial differential equations with constraints , numerical calculations are the realistic tools to study these equations for the actual physical events in the universe such as the gravitational waves . analyzing the einstein equations
numerically is called `` numerical relativity '' , these studies made a substantial contribution to observe the gravitational wave events .
these discoveries must open new physical fields to survey of lots of astrophysical events .
therefore , it is important to improve numerical techniques to make high accuracy numerical results .
although , there are some researches of the suitable numerical schemes of the einstein equations(e.g .
@xcite ) , it is not enough to perform precisely simulations .
however , it would be hardly to make best discretized equations of the einstein equations because of the strong nonlinearity .
in addition , some of the characteristics in the continuous system are often lost in the process of transforming the continuous equation to the discretized one .
there are some discretized schemes of preserving properties of the continuous equations .
one of the schemes is the discrete variational derivative method ( dvdm ) .
dvdm was proposed and has been extended by furihata , mori , matsuo , and yaguchi @xcite , and this method is one of the structure - preserving methods @xcite .
the method of formulating a discrete system using dvdm is similar to the variational principle . in the variational principle
, we provide a generator function such as the lagrange density and we can obtain a continuous system using the functional derivative of the function . on the other hand , in dvdm , we provide a discretized generator function of the system and we can obtain a discrete system using the discrete variational derivative . the crank - nicolson scheme ( cns ) and
the runge - kutta scheme are general methods of discretizing equations , these schemes often use in numerical relativity . on the other hand
, dvdm is a specific method of preserving constraints and diffusion characteristics in the continuous system .
since the einstein equations include the constraints , dvdm can provide a discretized einstein equations with constraint preserving properties . before we try to make the discretized einstein equations using dvdm , we studied the discretized maxwell s equations using the method because of the complexities of the einstein equations@xcite .
the maxwell s equations are classified the hyperbolic equations with constraints , the characters of the system are similar to those of the einstein equations . in @xcite
, the authors studied the numerical stabilities for the maxwell s equations using dvdm , and showed that dvdm is effective to perform stable simulations in the system . in this letter , indices such as @xmath0 run from @xmath1 to @xmath2 .
we use the einstein convention of summation of repeated up - down indices .
now , we review the processes of the standard method of discretized equations using general schemes and the new way using dvdm .
the discrete values of the variable @xmath3 are defined as @xmath4 .
the upper index @xmath5 and lower index @xmath6 denote the time components , and space components , respectively .
the details of this are in ref .
we adopt only the canonical formulation in this letter , so that these processes are only shown in the case that the generator function is the hamiltonian density .
first , we introduce the process of the general case of constructing the discrete equations such as cns and the runge - kutta scheme .
( a ) : we set a hamiltonian density @xmath7 with the variables @xmath8 and the canonical conjugate momenta @xmath9 .
( b ) : the dynamical equations are given as @xmath10 then we discretize the continuous equations following the ideas of the schemes such as cns and the runge - kutta scheme .
next , the processes of discrete equations using dvdm are as follows : ( a@xmath11 ) : we set a discrete hamiltonian density @xmath12 with the discrete variables @xmath13 and the discrete canonical conjugate momenta @xmath14 .
( b@xmath11 ) : the calculation of the discretized system using dvdm is @xmath15 where @xmath16 and @xmath17 are calculated as @xmath18
the lagrangian density of the einstein equations is given by @xmath20 where @xmath21 is lapse function , @xmath22 is the determinant of the three - metric @xmath23 , and @xmath24 is the three - scalar curvature .
@xmath25 is the extrinsic curvature , @xmath26 .
the word gr means the general relativity . with the variables @xmath23 and the canonical momenta @xmath27 , the hamiltonian formulation of the einstein equations can be constructed .
this formulation was given by arnowitt , deser , and misner ( adm ) , this is the first successful decomposition of the einstein equations to space and time @xcite .
if we discretize the equations of the formulation with dvdm , we face to two difficulties .
first is discretizing of @xmath22 .
the variational of @xmath22 in continuous case is @xmath28 it is hard to make the above relation in discretized case .
then , we use the conformal transformation such as @xmath29 to vanish @xmath22 in the lagrangian density . these relations are derived from following two conditions : ( i ) the lagrangian density @xmath30 does not include @xmath31 , ( ii ) the evolution equation of @xmath32 does not include @xmath31 .
if these conditions are satisfied , the @xmath31 is not also included in the expression of the hamiltonian density @xmath33 derived from the legendre transformation of the lagrangian density .
thus , the canonical formulation using @xmath33 does not include @xmath31 . with
, the lagrangian density can be expressed as @xmath34 where @xmath35 is the scalar curvature of @xmath32 , @xmath36 is the connection of @xmath32 , + @xmath37 , and @xmath38 .
note that the expressions of @xmath30 and @xmath39 are different , but they are same , that is to say @xmath40 .
the relation can be derived in straightforward .
then the hamiltonian density is defined as @xmath41 next is construction of suitable discretization of the @xmath24 .
this is defined as in continuous case such as @xmath42 note that there are the same relations between @xmath35 , @xmath36 , and @xmath32 .
this means that the conformal transformation does not overcome the difficulty of @xmath24 .
therefore we treat only @xmath19 to simplify @xmath24 in this letter .
if @xmath33 is the function of only @xmath19 , then the density can be expressed as @xmath43 the evolution equations are @xmath44 and the constraint equation is @xmath45 in the adm formulation , the constraint values are the hamiltonian constraint @xmath46 and the momentum constraints @xmath47 , thus there are four constraints .
however , there is only one equation of constraint in the above set of equations .
this is because we treat the case of @xmath33 as only @xmath19 function .
if we treat the case of @xmath48 as @xmath49 function , there are four constraint equations in the new formulation .
the evolution equation of can be derived as @xmath50 then the hamiltonian constraint is satisfied in the evolutions .
we call the evolution equation of constraint as constraint propagation ( cp ) .
we set the discrete hamiltonian density as @xmath51 then , the discretized evolution equations with dvdm are @xmath52 the discretized cp is derived as @xmath53 thus , the discretized cp of dvdm is the same as the continuous case . to compare with the numerical stability of the discrete system using dvdm , we select the crank - nicolson scheme ( cns ) which is used in the numerical relativity .
the discrete evolution equations with cns are @xmath54 and the discretized cp with the scheme is @xmath55 this is not equal zero in general because that the terms @xmath56 , @xmath57 , @xmath58 , and @xmath59 are not equal zero .
therefore , the constraint values would be changed of initial values using the cns .
we perform some simulations to study differences of the constraint violations using the two schemes .
then , we adopt the initial condition as @xmath60 this is the kasner solution which is one of the exact solutions of the einstein equations .
this solution expresses the expanding universe , and there is a singularity at @xmath61 .
then the initial conditions of @xmath21 , @xmath23 , and @xmath25 are given by @xmath62 and using the relations of becomes as @xmath63 in addition , it is hardly to make constraint violations in the above condition .
then , we add perturbations such as @xmath64 the other conditions in these simulations given by * initial time : @xmath65 . * time steps : @xmath66 .
* @xmath21(gauge condition ) : exact solution of . *
boundary condition : periodic .
there are @xmath67 time step variables in the discretized evolution equations using dvdm and cns ( - and - , respectively ) . in the simulations ,
the variables are calculated in iterative because of the nonlinearlity , then we set the numbers of times of iteration as fourth order in both schemes .
and horizontal axis means time .
the solid line is drawn by dvdm and the dotted line is by cns .
the line of cns is grow up after initial time . on the other hand ,
the one of dvdm is horizontal in the simulations .
[ fig : hamiltonain ] ] fig.[fig : hamiltonain ] is the constraint violations using dvdm and cns . in this figure ,
the solid line which is drawn by dvdm is horizontal , and the dotted line which is drawn by cns is rising .
this means that the values of the constraint @xmath48 of dvdm is unchanged in the evolution . on the other hand
, that of cns is increasing from the initial values .
therefore , these results are consistent with the results of analyzing cp in sec.[sec : discretizedeqs ] .
we derived suitable discretized equations of the einstein equations using the discrete variational derivative method ( dvdm ) . in the process of the discretization , there are two difficulties .
one is the existence of the determinant of @xmath23 , @xmath22 , in the lagrangian density , and the other is the complexity of the scalar curvature @xmath68 included in the density . for the existence of @xmath22 , we eliminated it by the conformal transformation of the variables . for the complexity of @xmath68 , we restricted the coordinate as only @xmath19
. then , we derived the discretized evolution equations of the variables and constraint in dvdm and the crank - nicolson scheme ( cns ) , showed the discrete constraint of dvdm is unchanged and the one of cns is changed .
in addition , we performed some simulations using the discretized equations of the dvdm and cns . then , the constraint values using the dvdm were conserved . on the other hand ,
the values using cns were not conserved .
these results are consistent with the analysis of the discretized evolution equations of the constraint using the dvdm and cns .
therefore the dvdm is the validity to perform stable simulations in the numerical relativity . in this letter
, we restricted the coordinates to simplify the complexities of the scalar curvature @xmath24 . to perform realistic simulations
, we should handle the @xmath24 in the coordinates @xmath49 .
so we would like to try it in the future work .
g. y. was partially supported by a waseda university grant for special research projects ( number 2016b-121 , 2016k-161 ) .
22 abbott , b. p. _ et al . _ ( ligo scientific collaboration ) , observation of gravitational waves from a binary black hole merger , phys
. rev . lett . * 116 * 061102 ( 2016 ) .
abbott , b. p. _ et al . _ ( ligo scientific collaboration and virgo collaboration ) , gw151226 : observation of gravitational waves from a 22-solar - mass binary black hole coalescence , phys .
lett . * 116 * 241103 ( 2016 ) .
r. richter and c. lubich , free and constrained symplectic integrators for numerical general relativity , class .
quantum grav . * 25 * 225018 , ( 2008 ) .
d. furihata and m. mori , a stable finite difference scheme for the cahn - hilliard equation based on the lyapunov functional , zamm z. angew .
mech . * 76 * , 405 - 406 ( 1996 ) .
d. furihata , finite difference schemes for @xmath69 that inherit energy conservation or dissipation property , j. comput . phys . *
156 * , 181 - 205 ( 1999 ) .
t. matsuo and d. furihata , dissipative or conservative finite difference schemes for complex - values nonlinear partial differential equations , j. comput.phys .
* 171 * , 425 - 447 ( 2001 ) .
d. furihata and t. matsuo , discrete variational derivative method , crc press / taylor & francis , london , ( 2010 ) .
y. aimoto , t. matsuo , and y. miyatake , a local discontinuous galerkin method based on variational structure , discrete contin .
* 8 * , 817 - 832 ( 2015 ) .
a. ishikawa and t. yaguchi , geometric investigation of the discrete gradient method for the webster equation with a weighted inner product , jsiam letters * 7 * , 17 - 20 ( 2015 ) .
h. hairer and c. lubich , geometric numerical integration , springer - verlag , berlin heidelberg , ( 2010 ) .
t. tsuchiya and g. yoneda , constraint - preserving scheme for maxwell s equations , arxiv : gr - qc/1610.04370 .
r. arnowitt , s. deser , and c. w. misner , in gravitation : an introduction to current research , edited by l. witten , wiley , new york , ( 1962 ) ; r. arnowitt , s. deser , and c. w. misner , republication of : the dynamics of general relativity , gen .
. gravit . * 40 * , 1997 - 2027 ( 2008 ) .
r. m. wald , general relativity , the university of the chicago press , chicago , ( 1984 ) . | we propose a new numerical scheme of evolution for the einstein equations using the discrete variational derivative method ( dvdm ) .
we derive the discrete evolution equation of the constraint using this scheme and show the constraint preserves in the discrete level .
in addition , to confirm the numerical stability using this scheme , we perform some numerical simulations by discretized equations with the crank - nicolson scheme and with the new scheme , and we find that the new discretized equations have better stability than that of the crank - nicolson scheme . |
the study of collisions between neutral atoms and molecules at low collision energies is a fast developing field in molecular physics @xcite .
this growing interest originates from the exotic scattering properties of molecules at low temperatures . at temperatures below @xmath3 10
k only a few partial waves contribute to the scattering , leading to dramatic changes in the dynamics .
scattering resonances can occur when the collision energy is degenerate with a bound state of the collision complex @xcite .
low collision energies also allow for external control over the collision dynamics by electromagnetic fields . at collision energies below a few kelvin ,
the perturbations due to the zeeman and stark effect become comparable to the translational energy , opening the possibility for controlled chemistry @xcite . in recent years
, a variety of novel experimental methods have been developed that enable scattering experiments at lower collision energies and/or with a higher precision than hitherto possible .
the buffer - gas cooling , the stark deceleration , and the velocity selection techniques have already successfully been applied to molecular scattering experiments @xcite . in the ultra - cold regime ,
spectacular advances have been made in the study of interactions between alkali - dimers near quantum degeneracy @xcite .
together with the collection of other techniques that are currently being developed , these methods have the potential to start a new era in molecular scattering experiments .
the stark deceleration technique has excellent potential for precise molecular scattering studies as a function of the collision energy . compared to conventional molecular beam sources
, a stark decelerator produces beams of molecules with a narrow velocity spread , perfect quantum state purity , and with a computer controlled velocity @xcite .
so far , two experimental approaches have been followed to use these monochromatic beams in molecular scattering studies . in 2006 ,
the first scattering experiment using a stark decelerated molecular beam was performed .
stark - decelerated oh radicals were scattered with a supersonic beam of xe atoms under 90 degree angle of incidence @xcite .
this crossed molecular beam configuration allowed the accurate measurement of the relative inelastic scattering cross sections as a function of the collision energy in the collision energy range of 50 to 400 @xmath2 .
recently , this experimental approach was improved significantly using a superior stark decelerator @xcite . with this decelerator
, scattering experiments can be performed with a better sensitivity , as has been demonstrated for the benchmark oh(@xmath4)-ar system @xcite . in both experiments ,
excellent agreement was obtained with cross sections determined by quantum close - coupling calculations based on high - quality _ ab initio _ oh - xe and oh - ar potential energy surfaces ( pes s ) . in another experiment , the approach to confine the stark - decelerated oh radicals in a permanent magnetic trap prior to the collision was followed @xcite .
collisions with the oh radicals were studied by sending supersonic beams of he atoms or d@xmath1 molecules through the trap .
information on the total collision cross sections could be inferred from the beam - induced trap loss that occurs through elastic as well as inelastic collisions .
the collision energy was varied from 60 to 230 @xmath2 for collisions with he and from 145 to 510 @xmath2 for collisions with d@xmath1 by changing the temperature of the pulsed solenoid valve used to produce the supersonic beams .
absolute collision cross sections were determined by calibrating the beam flux using a pressure measurement .
indications for quantum threshold scattering at a collision energy of 84 @xmath2 , equal to the energy splitting of the two lowest lying rotational levels of the oh radical , were found for collisions between oh and he . for oh - d@xmath1 collisions , a pronounced peak in the total cross section
was observed at collision energies around 305 @xmath2 , an energy that is equal to the energy splitting between the @xmath5 and the @xmath6 rotational levels of para - d@xmath1 .
the enhancement of the cross section at this energy was tentatively attributed to resonant energy transfer between the oh radicals and d@xmath1 molecules .
both experimental approaches have been successful in demonstrating the feasibility of using stark decelerated molecules in scattering experiments .
the experimental results of the latter experiment , however , could not be reproduced by theoretical calculations .
the decrease in the total cross section that was observed below 84 @xmath2 for the scattering between oh and he was not reproduced by rigorous quantum calculations for low temperature collisions of oh - he .
the calculations show that the total cross section _ increases _ significantly at collision energies below 100 @xmath2 @xcite .
the experimentally observed threshold behavior can be explained if the trap loss originates mainly from inelastic scattering , although this appears unlikely for the kinetic conditions of the experiment @xcite
. the measured cross section of @xmath3 100 @xmath7 at a collision energy of 150 @xmath2 is an order of magnitude larger than the theoretical total inelastic cross section , indicating that the majority of the collisions that lead to trap loss are indeed elastic .
the source of the discrepancy between theory and experiment , and whether or not the presence of the trapping field in the collision experiment can explain the observed cross sections , is at present unclear .
unfortunately , no rigorous quantum calculations are available for the oh - d@xmath1 system to compare calculated cross sections with experimental ones , or to investigate the physical origin of the intriguing peak that was observed at a collision energy of 305 @xmath2 . here
we complement the low - energy collision studies between oh - he and oh - d@xmath1 of sawyer _ et al . _
@xcite by investigating rotational energy transfer in collisions of stark - decelerated oh ( @xmath8 ) radicals with he atoms and d@xmath1 molecules in a crossed beam experiment under field free conditions .
the oh - he and oh - d@xmath1 center - of - mass collision energies are tuned from 120 @xmath2 to 400 @xmath2 and from 150 @xmath2 to 500 @xmath2 , respectively .
parity - resolved state - to - state relative inelastic scattering cross sections are accurately measured .
for the oh - he system , good agreement is obtained with the inelastic cross sections determined by close - coupled calculations based on the oh - he pes s used in ref .
@xcite , validating the theoretical predictions for the low - energy inelastic scattering between oh radicals and he atoms .
for the oh - d@xmath1 system , no strong variation in the state - to - state relative inelastic scattering cross sections is found at center - of - mass energies around 300 @xmath2 .
the interaction of oh radicals with , in particular , he atoms has been the subject of extensive experimental and theoretical investigations , and is of direct astrophysical relevance .
the rotational energy transfer of oh by collisions with h@xmath1 molecules in interstellar clouds is believed to play an important role in the formation of interstellar oh masers @xcite .
the ( mass scaled ) collision cross sections for the theoretically more tractable oh - he system are often used to model the collision dynamics in these environments . in a crossed beam experiment , schreel _ et al .
_ prepared the oh radicals in the upper @xmath9-doublet component of the @xmath10 level by hexapole state selection @xcite .
accurate state - to - state inelastic scattering cross sections were obtained at a collision energy of @xmath3 400 @xmath2 .
the effect of vibrational excitation of the oh radical on the rotational energy transfer has been investigated by wysong _
the bound states of the weakly bound oh - he complex were spectroscopically investigated by han and heaven @xcite .
the depolarization of rotationally excited oh radicals with he under thermal conditions has been studied using two - color laser spectroscopy by paterson _
all these experiments have been in good agreement with theoretical calculations based on accurate _ ab initio _ potential energy surfaces @xcite .
the interaction between oh radicals and d@xmath1 molecules has been studied less extensively . in crossed beam experiments ,
inelastic as well as reactive scattering processes have been studied at high collision energies @xcite .
the system has recently been treated theoretically using the quasi - classical trajectory method @xcite .
rotational inelastic cross sections for the related oh - h@xmath1 system have been calculated by van dishoeck and coworkers @xcite .
the experiments are performed in a crossed molecular beam machine that is schematically shown in figure [ fig : setup ] .
this machine has been used recently to study the rotational energy transfer in collisions between state - selected oh ( @xmath8 ) radicals and ar atoms as a function of the collision energy @xcite .
a detailed description of the production , stark deceleration and detection of the oh radicals , as well as of the procedure that is followed to tune the collision energy is given in ref .
@xcite ; we here limit ourselves to a brief summary
. a pulsed supersonic beam of oh radicals in the @xmath11 state is produced by photolysis of nitric acid seeded in an inert carrier gas .
the oh radicals that reside in the upper @xmath9-doublet component of @xmath12 parity are decelerated , guided , or accelerated with the use of a 2.6 meter long stark decelerator that is used in the so - called @xmath13 operation mode @xcite .
the oh radicals are scattered with a neat beam of he atoms or d@xmath1 molecules at a distance of 16.5 mm from the exit of the decelerator .
the beams scatter under 90@xmath14 angle of incidence and collisions take place in a field free region .
the density of the he and d@xmath1 molecular beams are kept sufficiently low to ensure single collision conditions . the beam of d@xmath1 molecules is produced using a gas of normal d@xmath1 , and both ortho - d@xmath1 and para - d@xmath1 molecules contribute therefore to the measured state - to - state inelastic cross sections . according to the statistical weights , 67% and 33% of the d@xmath1 molecules
are expected to reside in a rotational state belonging to ortho and para - d@xmath1 , respectively .
the rotational energy level diagram of d@xmath1 is schematically shown in figure [ fig : energy - level - scheme ] .
the energy level spacing between the lowest rotational states is large , and the d@xmath1 molecules are expected to predominantly reside in the @xmath15 and the @xmath5 levels under our experimental conditions .
the collision energy is varied by tuning the velocity of the oh radicals , and by choosing different temperatures for the solenoid valve used to produce the he and d@xmath1 beams .
the mean forward velocity of the he / d@xmath1 beam is measured by two microphone based beam detectors placed 300 mm apart . for a given temperature of the valve ,
similar beam speeds are measured for he and d@xmath1 as is to be expected for beam particles of identical mass .
the lowest collision energies are obtained when the solenoid valve is cooled to near liquid nitrogen temperatures , resulting in a minimum mean beam velocity of 996 m / s for he and 1042 m / s for d@xmath1 .
these velocities are expected for particles with a mass of 4 atomic units and near liquid nitrogen nozzle temperatures .
the slightly higher speeds that are measured for d@xmath1 beams are attributed to the extra degrees of freedom of the d@xmath1 rotor compared to he .
the lowest collision energy amounts to 120 @xmath2 and 150 @xmath2 for oh - he and oh - d@xmath1 , respectively . in the experiment by sawyer _
_ , a minimum collision energy of @xmath3 60 @xmath2 for oh - he and @xmath3 145 @xmath2 for oh - d@xmath1 was obtained @xcite .
it is noted that the significantly lower collision energy for oh - he that was reached in that experiment is not due to the use of trapped oh radicals .
a center of mass collision energy of @xmath3 60 @xmath2 for collisions between stationary oh radicals and helium atoms requires a he atom velocity of @xmath3 670 m / s .
this is much lower than the expected velocity for a supersonic beam of helium atoms at the temperature used , and the atomic beam that was employed by sawyer _
_ is believed to have been effusive - like @xcite .
collision energies up to 400 @xmath2 for oh - he and 500 @xmath2 for oh - d@xmath1 are reached by tuning the velocity of the oh radicals between 168 and 741 m / s , and by using temperatures of 293 k , 253 k , 213 k , 173 k , 133 k , and 93 k for the valve producing the he / d@xmath1 beam . the width ( full width at half maximum ) of the collision energy distribution depends on the collision energy , and ranges from @xmath3 20 @xmath2 at the lowest collision energies to @xmath3 60 @xmath2 at the highest collision energies .
collisional excitation of the oh radicals up to the @xmath16 and the @xmath17 state is measured .
these rotational states are schematically shown in the rotational energy level diagram in figure [ fig : energy - level - scheme ] , and are referred to hereafter as @xmath18 , where @xmath19 and @xmath20 are used to indicate the @xmath21 and the @xmath22 spin - orbit manifolds , respectively .
the inelastically scattered oh radicals are state - selectively detected via saturated laser induced fluorescence using different rotational transitions of the @xmath23 band .
the off - resonant fluorescence is collected at right angles and imaged into a photomultiplier tube ( pmt ) .
the diameter of the laser beam is approximately 8 mm , providing a detection volume that is larger than the intersection volume of both beams .
the experiment runs at a repetition rate of 10 hz . the beam that provides the collision partner
is operated every second shot of the experiment , and the collision signal results as the signal intensity difference of alternating shots of the experiment .
the collision energy is varied in a quasi - continuous cycle , as has been described in detail in ref .
for the strongest scattering channels , the fluorescence signals are recorded using an analog mode of detection ; the weak signals are analyzed using photon counting .
both modes of signal acquisition are calibrated with respect to each other .
to relate the measured signal intensities to collision induced populations , the different excitation rates for the different branches of the optical transitions used to probe the different rotational levels are taken into account in the data analysis @xcite .
the measured relative populations in the various rotational states directly reflect inelastic cross sections .
no density - to - flux transformation is required for crossed beam scattering experiments using a light particle as collision partner @xcite .
the validity of this assumption is verified by a measurement of the variation of the relative collision signals as a function of time in the overlapping beams .
no variation was recorded in the experiment , in agreement with model calculations of the detection probabilities of the scattered molecules .
we now describe the theoretical methods used to calculate the cross sections for the inelastic scattering between oh radicals and he atoms ; a discussion of the experimental relative cross sections is given in section [ sec : discussion ] .
fully quantum , close - coupling scattering calculations of inelastic collisions of oh radicals with he atoms have been performed recently by kos _
et al . _ in ref .
@xcite based on the rccsd(t ) potential energy surfaces of lee _ et al .
@xcite present cross sections for rotational transitions out of the @xmath24 state ; for the present experiment transitions out of the @xmath25 states are of relevance .
below we briefly describe the scattering methodology relevant for the oh - he system and some of the details of the calculations presented in ref .
@xcite .
the interaction between the open shell oh(@xmath26 ) radical and a spherical he atom is described by two pes s @xmath27 and @xmath28 , having @xmath29 and @xmath30 reflection symmetry in the plane containing the oh radical and the he atom @xcite .
the pes of @xmath29 and @xmath30 symmetry describes the oh - he interaction where the oh radical has its singly occupied @xmath31 orbital in and perpendicular to the triatomic plane , respectively . in scattering calculations
it is more convenient to construct the average potential @xmath32 and the half - difference potential @xmath33 of these pes s @xcite .
the hibridon program suite was used to carry out fully - quantum , close - coupling calculations of integral state - to - state scattering cross sections @xcite .
the channel basis was chosen to ensure convergence of the integral cross sections for all @xmath34 transitions with @xmath35 11.5 .
the calculated inelastic cross sections were converged to within 0.01 @xmath7 .
in figures [ fig : results_f1 ] and [ fig : results_f2 ] , the measured relative state - to - state inelastic scattering cross sections to levels in the @xmath36 manifold ( spin - orbit conserving collisions ) and the @xmath37 manifold ( spin - orbit changing collisions ) , respectively , are shown .
the cross sections that are obtained for the collision partners he and d@xmath1 are displayed in the left and right hand side of these figures . the scattering channels that correspond to excitation of the oh radicals to the two different @xmath9-doublet components of a given rotational state are grouped together . in the upper panels , the scattering channel that populates the @xmath24 state
is shown together with the channels that populate both @xmath9-doublet components of the @xmath38 rotational state . to facilitate a direct comparison between the scattering cross sections for oh - he and oh - d@xmath1 ,
identical axes are used in the panels that correspond to the same scattering channels .
the theoretically calculated cross sections for the scattering of oh with he , convoluted with the experimental energy resolution , are included as solid curves in the left panels . in the collision energy range that is probed ,
the rotational inelastic scattering of oh @xmath39 radicals with he atoms is dominated by excitation to the @xmath40 state .
the @xmath24 channel , corresponding to collisions that induce the @xmath41 @xmath9-doublet transition in the oh radical , appears rather weak .
this is in contrast with the scattering of oh radicals with ar and xe atoms for which the @xmath24 channel is the dominant inelastic scattering channel @xcite . the scattering channel populating the @xmath42 state appears exceptionally large , also at variance with the corresponding cross sections for the collision partners ar and xe . for spin - orbit manifold conserving collisions
, there is a strong propensity for final states of @xmath43 parity .
for spin - orbit manifold changing collisions populating the @xmath44 and @xmath45 states , very strong propensities are observed , showing a near symmetry selection rule .
collisions that populate the @xmath42 and @xmath46 states are approximately two orders of magnitude more effective than collisions populating the @xmath47 and @xmath48 states , respectively . at high collision energies , the relative state - to - state cross sections and the propensities
are consistent with the observations by schreel _
_ @xcite . in that experiment , however , the strong propensities were partially concealed due to a sizable initial population in the @xmath49 state @xcite .
the almost perfect quantum state purity of the packets of oh radicals that are used in the present experiment enables to unambiguously measure the cross sections of transitions to final states that are only weakly coupled to the @xmath25 initial state . throughout the range of collision energies ,
a good agreement is found with the computed cross sections .
the relative scattering cross sections and the propensities for transitions to one of the @xmath9-doublet components of the final rotational state , are reproduced well .
the largest difference between theory and experiment is found for the @xmath42 channel .
it is interesting to investigate the physical origin of the general behavior of the scattering cross sections and in particular the @xmath9-doublet propensities .
a general picture of the scattering of @xmath4 molecules with spherical objects has been developed by dagdigian _
@xcite , and has been applied to the inelastic scattering of oh radicals with atomic collision partners before @xcite . from the formal quantum analysis of the scattering , it follows that the coupling between rotational states can be evaluated from @xmath50\ ] ] where @xmath51 is the rotational quantum number of the oh radical , @xmath52 is the symmetry index of the rotational state , and @xmath53 is the projection of @xmath51 onto the internuclear axis . primed quantum numbers indicate the post - collision conditions .
the terms @xmath54 and @xmath55 are the expansion coefficients of the average and difference potentials in terms of regular and associated legendre polynomials , respectively .
the sum in equation ( [ eq : matrixelement ] ) is performed over the expansion index @xmath56 .
the factor @xmath57 is only nonzero for states that are coupled by the interaction potential , and needs for our analysis no further discussion .
essential in the understanding of the inelastic cross sections are the values for the @xmath54 and the @xmath55 coefficients , and the role of the @xmath58 and @xmath59 factors . both @xmath58 and @xmath59 are independent of the interaction potential , and are determined exclusively by the rotational energy level structure of the molecule .
the values of @xmath58 and @xmath59 for oh radicals in the @xmath8 level are tabulated in ref .
@xcite . for a pure hund s case ( a ) molecule , the values for @xmath59 are zero for spin - orbit manifold conserving collisions
, whereas the factors @xmath58 are zero for spin - orbit manifold changing collisions
. consequently , spin - orbit conserving and spin - orbit changing transitions are induced exclusively by @xmath60 and @xmath61 , respectively . within each manifold , @xmath62 parity changing collisions are governed by the terms for which @xmath63 odd , while @xmath62 parity conserving collisions are described by the @xmath64 even terms .
the propensities for preferred excitation to the @xmath43 or @xmath12 component of a final rotational state originate from the different values for the relevant products @xmath65 or @xmath66 that govern these transitions . for molecules like oh that can not be described by a pure hund s case ( a ) coupling scheme , both the factors @xmath58 and @xmath59 are nonzero , and interference between the average and difference potentials occurs .
the @xmath67 electronic ground state of the oh radical originates from a @xmath68 electron occupancy , leading to predominantly positive values for @xmath60 and @xmath61 @xcite . as a result , final states for which the dominant @xmath58 and @xmath59 factors have equal signs are coupled more strongly to the @xmath25 initial state than final states for which @xmath58 and @xmath59 have opposite signs .
the former is the case for the final states of @xmath43 parity in the @xmath36 spin - orbit manifold , and for final states of @xmath12 parity in the @xmath37 manifold , contributing to the observed propensities .
the anomalous propensity that is observed for the @xmath42 state can be understood from equation ( [ eq : matrixelement ] ) .
excitation into the @xmath47 state is governed exclusively by @xmath69 in combination with a small value for @xmath70 , while the @xmath71 term in combination with a large value for @xmath72 dominates the excitation into the @xmath42 state @xcite .
the measured relative state - to - state cross sections directly yield qualitative information on the expansion coefficients of the potential energy surfaces that govern the scattering between oh and he .
collisions that populate the @xmath24 and @xmath49 states are governed by the coefficients for which @xmath73 odd , while the cross sections for excitation to the @xmath40 state is governed by the @xmath73 even coefficients .
the observed ratio of the state - to - state cross sections indicates that the leading @xmath73 even terms @xmath74 , @xmath75 , and @xmath71 contribute significantly to the interaction potential .
the relatively large cross section for spin - orbit changing collisions populating the @xmath42 state , as well as the strong propensities that are generally observed , suggests that the @xmath71 coefficient of the difference potential plays a significant role in the scattering between oh radicals and he atoms .
these effects can be rationalized by the nature of the oh - he interaction potential .
the interaction between oh and he is rather weak and the anisotropy of the potential energy surface is small .
the @xmath28 pes has two shallow and almost equally deep potential wells for the collinear oh - he and the ho - he geometry , with well depths of 27 @xmath2 and 22 @xmath2 , respectively @xcite .
consequently , the @xmath73 even coefficients that describe the head - tail symmetric parts of the potential energy surfaces contribute significantly to the scattering .
the inelastic scattering of oh radicals with d@xmath1 molecules shows interesting differences compared to oh - he .
the largest cross section is observed for collisions that populate the @xmath24 state . for the channels that populate the @xmath38 states ,
a propensity for the @xmath9-doublet component of @xmath43 symmetry is observed for low collision energies , that vanishes for collision energies of about 500 @xmath2 .
the other scattering channels show only modest propensities .
the spin - orbit changing collisions appear rather weak .
although the formalism that was applied above to the scattering of oh with he does not strictly apply to non - spherical collision partners , we can use the formalism to obtain a physical interpretation of the differences between the scattering of oh with he and d@xmath1 .
this comparison is particularly interesting , as both collision partners have equal mass , and mass effects in the dynamics cancel .
the interaction of d@xmath1 molecules with oh radicals is stronger and induction forces are more important than for the interaction between oh and he .
this suggests that the pes is less head - tail symmetric in comparison to oh - he , and the coefficients of the potential for which @xmath73 odd gain importance compared to the @xmath73 even coefficients .
this results in a larger cross section for @xmath9-doublet changing collisions populating the @xmath24 state , smaller spin - orbit changing transitions , and less pronounced propensities for preferred excitation to one of the two components of a @xmath9-doublet .
similar effects have been observed in state - to - state inelastic scattering experiments of oh radicals with polar collision partners such as co@xmath1 and hcl @xcite . in the relative inelastic scattering
cross sections , no effect is seen from the internal rotational degrees of freedom of the d@xmath1 molecule .
in particular , no strong variation of the cross sections at collision energies around 300 @xmath2 is observed , that could be indicative of resonant energy transfer between the oh and the d@xmath1 rotors @xcite .
we have presented measurements of the state - to - state rotational inelastic scattering of stark - decelerated oh ( @xmath76 ) radicals with he atoms and d@xmath1 molecules in the 100 - 500 @xmath2 collision energy range .
the collision energy dependence of the relative inelastic scattering cross sections , the threshold behavior of inelastic channels , and the energy dependence of the state - resolved propensities are accurately determined . for the scattering of oh with he , good agreement is found with the inelastic scattering cross sections determined from quantum close - coupled scattering calculations based on high - quality _ ab initio _ oh - he pes s .
the almost perfect quantum state purity of the stark - decelerated packets of oh radicals eliminates the contamination of the scattering signals by initial populations in excited rotational states .
this facilitates a quantitative study of collision induced transitions to states that are only weakly coupled to the initial state , and enables the observation of the exceptionally strong propensities for the inelastic scattering between oh radicals and he atoms .
the genuine relative state - to - state inelastic scattering cross sections that are measured allow for a more accurate comparison with computed cross sections than hitherto possible .
significant differences are found between the inelastic scattering of oh - he and oh - d@xmath1 .
although no rigorous quantum calculations have been performed for oh - d@xmath1 , these differences can be understood from the different nature of the oh - d@xmath1 interaction potential .
no effect of the rotational degrees of freedom of the d@xmath77 molecule has been observed in the relative inelastic scattering cross sections . our measurements on the low - energy scattering between oh radicals and he / d@xmath1 complement recent scattering experiments in which pulsed beams of he / d@xmath1 are directed through a sample of magnetically trapped oh radicals .
in that experiment , inelastic collisions , as well as elastic collisions that impart a sufficient amount of kinetic energy to the oh radicals , lead to a reduced number of oh radicals in the trap .
recent quantum scattering calculations , based on the same pes s as used in the present work , have not been able to reproduce the trap loss observations for oh - he , however .
the state - to - state experiments reported in the present article are not sensitive to elastic scattering , but validate the field - free inelastic scattering cross sections for oh - he determined from quantum scattering calculations .
the theoretical description of low - energy inelastic collisions of these elementary systems is thus adequate ; more experimental and theoretical work is needed to find the source of the discrepancy between theory and beam - trap experiments . for the accurate interpretation of collision experiments using trapped molecules ,
it is essential to establish the role of elastic scattering and the influence of the trapping potential on the measured trap loss .
this work is supported by the esf euroquam programme , and is part of the copomol ( collisions of cold polar molecules ) project . the expert technical assistance of henrik haak , georg hammer , and the fhi mechanical and electronics workshops are gratefully acknowledged .
we thank timur tscherbul for stimulating discussions .
the theoretical calculations were supported by the u. s. national science foundation , grant number che-0848110 . | we present an experimental study on the rotational inelastic scattering of oh ( @xmath0 ) radicals with he and d@xmath1 at collision energies between 100 and 500 @xmath2 in a crossed beam experiment .
the oh radicals are state selected and velocity tuned using a stark decelerator .
relative parity - resolved state - to - state inelastic scattering cross sections are accurately determined .
these experiments complement recent low - energy collision studies between trapped oh radicals and beams of he and d@xmath1 that are sensitive to the total ( elastic and inelastic ) cross sections ( sawyer _ et al .
_ , _ phys .
rev .
lett . _ * 2008 * , _ 101 _ , 203203 ) , but for which the measured cross sections could not be reproduced by theoretical calculations ( pavlovic _ et al .
_ , _ j. phys .
chem .
a _ * 2009 * , _ 113 _ , 14670 ) .
for the oh - he system , our experiments validate the inelastic cross sections determined from rigorous quantum calculations . |
spin - orbit interaction influences the electronic and transport properties of semiconductors . for example
, it lifts the degeneracy of the valence bands , modifies the electron g - factor,@xcite and causes skew scattering in the presence of ( spinless ) impurities .
such a skew scattering is a possible mechanism for the extrinsic spin hall effect.@xcite in addition , spin - orbit interaction plays an important role in the recently proposed intrinsic spin hall effect.@xcite it is also crucial in the mechanisms of spin relaxation and optical orientation in semiconductors.@xcite it is highly desirable to generate flows of polarized spins in semiconductors with the help of the spin - orbit interaction . in these endeavors ,
the rashba spin - orbit coupling@xcite in two - dimensional electron gas ( 2deg ) plays a special role since it allows manipulation of spin flows by varying the gate bias .
this has motivated several creative proposals for its application.@xcite to explore the possibilities , the effect of rashba coupling in many types of mesoscopic structure have been investigated , such as a quantum wire,@xcite a quantum ring,@xcite and a quantum dot.@xcite in several studies , it was found that a device with a simple geometry , in combination with the rashba coupling , could serve as a spin filter .
for example , the device could be a t - shaped channel,@xcite a quantum point contact,@xcite parallel interfaces that cause double refraction,@xcite or even just a curved wire.@xcite after applying a magnetic field , we could further build a spin filter based on electron focusing,@xcite or based on the an interferometer of the stern - gerlach type.@xcite in this paper , we study the scattering of electrons by a disk in the 2deg with rashba coupling.@xcite the analysis can be applied to a wide range of situations in which the radius @xmath0 of the disk can be much smaller , roughly the same , or much larger than the electron wave length @xmath1 .
for example , the usual impurity scattering can be simulated with @xmath2 , while the scattering by an artificial mesoscopic disk corresponds to @xmath3 . here
we focus on the latter case to search for possible effect of local spin generation .
we find that , because of the rashba coupling , spin - up and spin - down electrons are indeed separated and accumulate in regions of curved stripes .
the associated charge and spin currents are analyzed in details .
in addition , the properties of symmetry , as well as the differential cross - section , are also investigated.@xcite this paper is organized as follows : sec .
ii is the theoretical analysis . in sec .
iii , the results from numerical calculations are presented .
iv is the conclusion .
consider a two - dimensional electron system with a circular disk at the origin , @xmath4 where @xmath5 when @xmath6 and 0 otherwise . in the following
, the potential of the disk is considered infinite ( hard " disk ) .
the two - component wave function @xmath7 with energy @xmath8 satisfies the coupled equations : @xmath9 it follows that the energy eigenstates of the hamiltonian with momentum @xmath10 , helicity @xmath11 ( @xmath12 ) , and angular momentum @xmath13 ( where @xmath14 is an integer ) are @xmath15 where @xmath16 can be a bessel function , a neumann function , or their linear combination , such as a hankel function
. we will choose the hankel functions as the eigen - basis since their behavior at large radius suits the boundary condition for scattering .
the energy eigenstate of the schroedinger equation with energy @xmath8 and angular momentum @xmath13 can be written as @xmath17 where @xmath18 .
the first two terms have positive helicity ; while the other two terms have negative helicity .
the most general eigenstate of the schroedinger equation with energy @xmath8 is a superposition of the @xmath19 s , where the coefficients @xmath20 , and @xmath21 are determined by boundary conditions . because of the circular symmetry of the potential , the angular momentum is conserved for each @xmath14-component .
therefore , each component can be considered independent during the scattering .
the incident plane wave with momentum @xmath10 and helicity @xmath11 can be decomposed as the following linear superposition : @xmath22 the bessel functions @xmath23 can be further decomposed as hankel functions @xmath24 and @xmath25 , the former correspond to outgoing circular waves , while the later correspond to incoming circular waves ( with no phase shift ) .
if an incident wave has a definite helicity , @xmath26 , then by comparing eq .
( 4 ) with the components of eq . ( 5 ) at large distance , one will obtain @xmath27 and @xmath28 ( i.e. no incoming circular wave with negative helicity ) for all @xmath14 .
the coefficients @xmath29 and @xmath30 need to be determined from the boundary condition at @xmath31 . for a hard disk , it can be shown that @xmath32 where @xmath33 and @xmath34 .
notice that the nonzero probability amplitudes @xmath30 lead to outgoing waves with flipped helicity . for reference , if the incident wave is @xmath35 , then @xmath36 and @xmath37 .
at the mean time , the roles of @xmath10 and @xmath38 , as well as the roles of @xmath29 and @xmath30 , have to be interchanged . a note on the unitary condition : for convenience of discussion , consider an incoming wave with positive helicity and angular momentum @xmath13 ( i.e. , the @xmath39-wavelet ) .
because the angular momentum is conserved during the scattering , the electron can only be scattered to @xmath29 and @xmath30 channels with the same @xmath14 . from particle conservation at large distance ,
one expects that the probability amplitudes in eq .
( 4 ) should satisfy @xmath40 for all @xmath14 , which has indeed been confirmed in our numerical calculation .
the system has a mirror symmetry with respect to the @xmath41-axis .
therefore , by analyzing the schrodinger equation with @xmath42 replaced by @xmath43 , one finds @xmath44 , where @xmath45 , and @xmath46 is the mirror - reflected point of @xmath47 . such as relation
can also be obtained by a space inversion of the ( three - dimensional ) coordinate , followed by a rotation with respect to the new @xmath42-axis by 180 degrees .
consequently , for the expectation value of the spin , we have @xmath48 in a rashba system , the current density operator is defined as , @xmath49 therefore , the distribution of the expectation value of the current density has the following symmetry : @xmath50 we adopt the generally accepted definition of the spin current density operator,@xcite @xmath51 , whose expectation values have the symmetries , @xmath52 these symmetries will be confirmed by the numerical results in sec .
iii . for convenience
, the wave function @xmath53 can be seperated into an incident plane wave and a scattered wave . at large distance with @xmath54 ,
the scattered wave has the asymptotic form , @xmath55 where @xmath56 it can be shown that @xmath57 .
also , for the incoming plane wave @xmath58 , @xmath59 therefore , @xmath60-spinor and @xmath61-spinor possess spins with opposite directions at large distance .
both spins lie on the plane and are perpendicular to the direction of propagation .
one can obtain the same equations for the incoming wave @xmath62 , but the signs of the spin expectation values are opposite .
after a straightforward calculation , one can show that the scattered current density at large distance is @xmath63 from which the differential cross - section @xmath64 can be calculated . for incoming waves @xmath26 and @xmath35 , the current densities
@xmath65 and @xmath66 are @xmath67 and @xmath68 respectively .
in fact , they are equal in magnitude if the two incident waves have the same energy . therefore , the differential cross - sections @xmath69 for incoming waves with helicity @xmath11 are @xmath70 where @xmath71 is the @xmath60-spinor in eq .
( 12 ) , but with the coefficients @xmath29 and @xmath39 suitably chosen for the scattering of @xmath72 , similarly for @xmath73 ( see the discussion following eq . ( 6 ) ) .
the differential cross - sections @xmath74 and @xmath75 represent helicity - preserved and helicity - flipped scatterings respectively .
if the incoming wave is an incoherent mixture of both helicities with fractional populations @xmath76 , then the differential cross - section is simply the weighted average of the two differential cross - sections : @xmath77 . , with @xmath78 .
the incident wave is an equal mixture of both helicities @xmath79 .
lighter ( darker ) regions represent the regions of spin up ( down ) . the peak intensity in this figure is approximately 0.22 ( 1 for fully polarized).,width=268 ]
in the following , we report on the distributions of spin density , charge current density , and spin current density , assuming the wave length of the incident wave @xmath80 .
we have also studied the cases with a larger @xmath1 ( e.g. @xmath81 ) and a smaller @xmath1 ( e.g. @xmath82 ) .
these results are not presented since the main difference is the change of scales . in the limit of @xmath83 ,
whose scale is more relevant to the case of impurity scattering , only the components with the smallest angular momentum ( @xmath14=0 and @xmath84 ) need to be considered . from eq .
( [ ac ] ) , one finds that both @xmath85 and @xmath86 as @xmath87 .
therefore , there would be little change of helicity in the long wave length limit . for comparison with realistic values ,
we choose @xmath88 for electrons in the gaas - algaas heterojunctions .
the radius of the disk is fixed at @xmath0=1000 @xmath89 . the corresponding fermi energy and electron density for @xmath90 are 2.2 mev and @xmath91/@xmath92 respectively , which are typical values . to enhance the visual effect of the spin - orbit coupling , the rashba energy @xmath93 is chosen to be one - tenth of the kinetic energy @xmath94 , which requires @xmath95 nm - ev , about one order of magnitude larger than the value in gaas . with wavelength @xmath80 .
[ @xmath78],width=268 ] the out - of - plane ( @xmath97 ) spin component results from a spin - unpolarized incident wave with an equal ( incoherent ) mixture of both helicities is plotted in fig . 1 , which is antisymmetric with respect to the horizontal @xmath41-axis : @xmath98 .
notice that the incident waves with opposite helicities ( but the same energy ) have different wave vectors ( @xmath99 ) .
therefore , their interference patterns for opposite helicities with spins point at opposite directions are slightly displaced with respect to each other . because of
such a displacement between @xmath100 and @xmath101 , regions with net @xmath97-spin still exist after partial cancellation .
the existence of @xmath102 relies on the scattered part of the wave function in the near - field region [ see eqs .
( [ scattered ] ) and ( [ far ] ) ] and would decay to zero at large distance . unlike the spins in fig .
1 , the distribution of charge current ( fig .
2 ) , as well as the differential cross - section ( fig .
3 ) , which is defined using the _ charge _ current densities , are not sensitive to the strength of the rashba coupling , and look very similar even if the coupling is turned off . notice that rashba spin - orbit coupling in fact preserves the helicity .
the major cause of the helicity flip is the potential @xmath103 , which is incompatible with the helicity operator . in fig . 2
, the distribution of charge current density from the scattering of @xmath26 shows the expected pattern of the flow around the disk .
if the incident wave is a mixed state , then instead of cancellation , the slightly displaced current densities from both helicities will add up . ) for wave length @xmath80 ( solid line for @xmath104 , dashed line for @xmath105 ) and @xmath106 ( dash - dotted line for @xmath104 , dotted line for @xmath105 ) .
the numbers in the legend are total cross - sections after integration over angle .
[ @xmath78],width=268 ] the differential cross - sections for @xmath26 are shown in fig .
it can be seen that the helicity - preserved scattering ( @xmath104 ) peaks at the forward direction ( @xmath107 ) , while the helicity - flipped scattering ( @xmath105 ) peaks at the backward direction ( @xmath108 ) . at the backward direction , @xmath109 .
therefore , the helicity of the electron has to be flipped , but _ its spin remains conserved_. at longer wave length @xmath106 , @xmath105 gains more weight and
the total differential cross - section @xmath110 becomes more isotropic .
, @xmath111 , and @xmath112 .
the incident wave has a well - defined helicity @xmath100 with @xmath113 .
[ @xmath78 ] ] the distributions of spin current density are shown in fig .
4.@xcite a prominent feature in the figures is the overall trend for the @xmath114 vectors to point to the left .
this counter - intuitive behavior is simply related to the fact that the spin current density is equal to the product of velocity and spin , where the velocity is in the positive @xmath41-direction and the spin points to the _ minus _
@xmath42-direction for @xmath26 , therefore the scattering @xmath115 vectors generically point to the left . on the contrary , for @xmath35 with spin points to the _ positive _ @xmath42-direction
, the direction of the flow will be reversed .
this also explains why the magnitudes of the spin currents @xmath116 are small in most of the regions , since the original incident current has no @xmath117 and @xmath102 components . in fig .
4 , it can be seen that @xmath118 oscillates both in amplitude and direction between the curved stripes .
an incident wave with opposite helicity would reverse such a flow .
therefore , part of these flows are cancelled if the incident current is not polarized . however , local spin current that oscillates in space still exist , similar to the case of the spin density in fig . 1 .
all of the local spin currents @xmath119 would vanish for unpolarized incident electrons once the rashba coupling is turned off .
however , if @xmath120 is nonzero but the disk is removed ( i.e. , free space ) , then there exist spin currents @xmath121 , @xmath122 , and @xmath123 , which is the background spin current cautioned by rashba.@xcite it reflects the unsatisfying current status on a proper definition of the spin current .
our system with unilateral current flow seems to be the same as the 2deg driven by an electric field,@xcite but there is no global transverse spin current @xmath118 , no matter the disk is removed or not .
this does not contradict the result of the proposed intrinsic spin hall effect in a clean 2deg.@xcite the incident charge current flowing to the right can be understood as originating from the slightly unbalanced electrochemical potentials on the two leads far away . in our case , all the electrons are moving along the direction of the potential gradient , instead of moving at all directions on the fermi surface in sinova _
s paper.@xcite consequently , no spin hall effect is expected if one follows similar semiclassical analysis in ref . .
in summary , the influence of the spin - orbit coupling on the scattering of 2d electrons from a hard disk is studied .
such a simple setup offers us a good opportunity to investigate the properties of spin and spin current in details .
we focus our attention on the near - field regime , where the scattered wave is comparable to the incident wave , and appreciable out - of - plane spins can be found .
this work offers us a clear understanding of the microscopic dynamics around the mesoscopic disk , and could serve as a basis for future works considering a threaded magnetic flux in the disk , or a hybrid device involving a disk as a component . , by c. kittel , wiley & sons , 1963 .
hirsch , phys .
lett . * 83 * , 1834 ( 1999 ) .
s. murakami , n. nagaosa , and s. c. zhang , science * 301 * , 1348 ( 2003 ) ; s. murakami , cond - mat/0405003 .
jairo sinova , dimitrie culcer , q. niu , n. a. sinitsyn , t. jungwirth , and a. h. macdonald , phys .
lett . * 92 * , 126603 ( 2004 ) .
igor zutic , jaroslav fabian , s. das sarma , rev .
mod . phys .
* 76 * , 323 - 410 ( 2004 ) ; w. zawadzki and p. pfeffer , semicond .
* 19 * , r1 ( 2004 ) . , edited by f. meier and b.p .
zakharchenya , north - holland , 1984 .
e. i. rashba , sov .
solid state * 2 * , 1224 ( 1960 ) ; y. a. bychkov and e. i. rashba , jetp lett . * 39 * , 78 ( 1984 ) .
for example , the current modulator proposed by s. datta and b. das , appl .
. lett . * 56 * , 665 ( 1990 ) ; marco g. pala , michele governale , jurgen konig , ulrich z@xmath124licke , and giuseppe iannaccone , phys .
b * 69 * , 045304 ( 2004 ) .
m. governale , u. z@xmath124licke , phys .
b * 66 * , 073311 ( 2002 ) ; m. governale , u. z@xmath124licke , cond - mat/0407036 . j. splettstoesser , m. governale , u. z@xmath124licke , phys . rev .
b * 68 * , 165341 ( 2003 ) ; s.q . shen , z.j .
li , and z. ma , app .
. lett . * 84 * , 996 ( 2004 ) ; s. souma and b.k .
nikoli@xmath125 , phys . rev .
b * 70 * , 195346 ( 2004 ) ; g.s .
lozanno and m.j .
s@xmath126nchez , cond - mat/0505394 .
halperin , ady stern , yuval oreg , j.n.h.j .
cremers , j.a .
folk , and c.m .
marcus , phys .
lett . * 86 * , 2106 ( 2001 ) ; w.h kuan , c.s .
tang , w. xu , cond - mat/0403098 ; e. tsitsishvili , g.s .
lozano , and a.o .
gogolin , phys .
b * 70 * , 115316 ( 2004 ) . a short but insightful review on spintronics can be found in m.i .
dyakonov , cond - mat/0401369 .
a.a . kiselev and k.w .
kim , appl .
. lett . * 78 * , 775 ( 2001 ) .
ramaglia , d. bercioux , v. cataudella , g. de filippis , c.a .
perroni , and f. ventriglia eur .
j. b * 36 * , 365 ( 2003 ) . v. m. ramaglia , d. bercioux , v .
cataudella , g. de filippis and c. a. perroni , j. phys . : condens .
matter * 16 * , 9143 ( 2004 ) .
trushin and a.l .
chudnovskiy , cond - mat/0505104 .
gonzalo usaj and c. a. balseiro , phys .
b * 70 * , 041301 ( r ) ( 2004 ) ; l.p .
rokhinson , y.b .
lyanda - geller , l.n .
pfeiffer , k.w .
west , phys .
lett . * 93 * , 146601 ( 2004 ) .
r. ionicioiu and i. damico , phys .
b * 67 * , 041307(r ) ( 2003 ) . in the absence of spin - orbit coupling , the study of two - dimensioanl electron scattering by a disk can be found in i.r .
lapidus , am .
* 50 * , 45 ( 1982 ) ; p.a .
maurone and c. shiomos , _ ibid_. * 51 * ( 1983 ) ; i.r .
lapidus , _ ibid_. * 54 * ( 1986 ) .
this report is abridged from jr - yu yeh s master thesis ( july , 2005 , national taiwan normal univ . ) , in which the distributions of charge , spin , charge current , and spin current are studied using various ratios of @xmath127 .
see , for example , s.i .
erlingsson , j. schliemann , d. loss , phys .
b * 71 * , 35319 ( 2005 ) ; e.i .
rashba , cond - mat/0404723 .
we have also plotted the same distributions using the spin current recently proposed by ping zhang , junren shi , di xiao , and qian niu , cond - mat/0503505 . for @xmath128 and @xmath115 ,
the patterns of the variation look similar to the ones shown in fig . 4 ; for @xmath118 , the stripes on the left of the disk are enhanced and more visible . e.i .
rashba , cond - mat/0408119 , invited talk at the 3rd intern .
conf . on physics and applications of spin - related phenomena in semiconductors ,
santa barbara ( ca ) , july 21 - 23 . | electrons with spin - orbit coupling moving in mesoscopic structures can often exhibit local spin polarization . in this paper , we study the influence of the rashba coupling on the scattering of two - dimensional electrons from a circular disk .
it is observed that spin - polarized regions exist , even if the incident electrons are unpolarized .
in addition to the distributions of charge and spin current in the near - field region , we also analyze the symmetry and the differential cross - section of the scattering . |
in 1992 , a numerical investigation of the equation of state ( eos ) of the st2 model @xcite in the supercooled region suggested the possibility of a liquid - liquid ( ll ) critical point in water @xcite . this initial study has subsequently generated a large amount of numerical and experimental work @xcite .
in addition to the conceptual novelty of a one - component system with more than one liquid phase , the existence of the associated ll critical point can also rationalize many of the thermodynamic anomalies which characterize liquid water ( e.g. the density maximum and compressibility minimum ) , and which become more pronounced in the supercooled regime .
furthermore , the existence of two distinct liquid phases of supercooled water can explain the polyamorphism which characterizes the glassy phase @xcite .
indeed , simulations suggest that the low density amorphous solid form of water is similar to the structure of the low density liquid ( ldl ) phase , while the relaxed very - high density amorphous solid is related to the high density liquid ( hdl)@xcite .
evidence in support of a liquid - liquid critical point in water , and in other liquids with tetrahedral structure , has increased over time .
a number of classical models for water , including the recently developed and optimized tip4p/2005 @xcite , exhibit a van der waals inflection in their eos at low temperature @xmath0 that is evidence of phase coexistence between two liquid states @xcite .
the occurrence of a ll transition has also been proposed for silica @xcite , and more recently , evidence for a ll critical point and its associated thermodynamic anomalies have been presented for the stillinger - weber model of silicon @xcite . indeed , it is notable that the most compelling evidence for ll critical points has been generated _ in silico _ @xcite . in almost all cases ,
ll critical points are predicted to occur in deeply supercooled liquids , where crystallization ( in experiments ) has so far prevented direct observation of such phenomenon in bulk systems .
compared to experiments , ll phase transitions are more readily observed in numerical studies because heterogeneous nucleation is not a factor , and the small system size ( usually less than one thousand molecules ) decreases the probability of observing the appearance of a critical crystal nucleus in the simulation box on the time scale of typical simulations .
computer simulations have thus allowed the study of the liquid eos under deeply supercooled conditions , on time scales longer than the structural relaxation time of the liquid but smaller than the homogenous nucleation time . under these conditions ,
equilibrium within the metastable basin of the liquid free energy surface can be achieved without interference from crystal nucleation processes .
nonetheless , evaluations of the eos via simulations in the canonical ensemble , or at constant pressure , do not provide a way to accurately estimate the location of the ll critical point found in water models , or to determine its universality class . only recently , in 2009 , liu , et al .
@xcite reported the first numerical investigation of st2 water in the ll critical region , performing simulations in the grand canonical ensemble for different values of @xmath0 and of the chemical potential @xmath1 , and implementing ewald sums to account for the long - range contributions to the electrostatic interactions . in this important contribution , the authors provided for the first time evidence of a density of states that is a bimodal function of the density @xmath2 , a necessary feature of a ll critical point .
importantly , the authors also showed that the fluctuations of the order parameter ( a combination of density and energy ) are consistent with the expected shape for a critical system in the ising universality class .
more recently , limmer and chandler @xcite have questioned the interpretation of all previously published simulations based on the st2 potential , arguing that `` that behaviors others have attributed to a liquid - liquid transition in water and related systems are in fact reflections of transitions between liquid and crystal . '' in the case of the recent calculations of liu , et al .
@xcite , ref .
@xcite proposes that `` the liu et al .
result is a non - equilibrium phenomenon , where a long molecular dynamics run initiated from their low - density amorphous phase and run at constant @xmath0 and @xmath3 will eventually equilibrate in either the low density crystal or ( more likely ) in the higher density metastable liquid . ''
it is thus of paramount importance to independently check the findings of liu , et al .
@xcite , and at the same time , test whether or not the ldl phase is truly a disordered liquid phase characterized by a well - defined metastable equilibrium that is distinct from the crystal phase . in this article , we conduct these tests by carrying out an independent evaluation of the density of states based on the successive umbrella sampling technique @xcite .
we implement the original st2 model , with the reaction field correction for the long range electrostatic forces , rather than ewald sums , to be able to strictly compare our results with previously published data for st2 @xcite , as well as to test if the ll transition is robust and independent of the treatment of the long range interactions . as we show below
, we find that our results are entirely consistent with those of liu , et al .
@xcite , as well as with earlier simulation data .
we further find that there is no contribution to the density of states due to crystal formation , confirming the distinct existence of both the hdl and ldl phases for @xmath0 less than @xmath4 , the temperature of the ll critical point .
we study the original st2 potential as defined by rahman and stillinger @xcite , with reaction field corrections to approximate the long - range contributions to the electrostatic interactions . in the st2 potential ,
water is modeled as a rigid body with an oxygen atom at the center and four charges , two positive and two negative , located at the vertices of a tetrahedron .
the distances from the oxygen to the positive and negative charges are 0.1 and 0.08 nm , respectively .
the oxygen - oxygen interaction is modeled using a lennard - jones potential with @xmath5 nm and @xmath6 kj / mol .
we truncate this lennard - jones interaction at @xmath7 , accounting for the residual interactions through standard long range corrections , i.e. assuming the radial distribution function can be approximated by unity beyond the cutoff .
the charge - charge interactions are smoothly switched off both at small and large distances via a tapering function , as in the original version of the model @xcite .
our grand canonical monte carlo ( mc ) algorithm is based on roto - translational moves , insertions , and deletions , each attempted with ratios 2:1:1 .
our simulation box is cubic with sides of length @xmath8 nm .
the displacement move is accomplished by a random translation in each direction of up to @xmath9 nm and a random rotation of up to @xmath10 rad around a random direction , resulting in an acceptance ration of about 50@xmath11 .
insertion and deletion moves have a much smaller acceptance ratio , of the order of @xmath12 .
the simulations have thus been performed for more than @xmath13 attempted insertion / deletion moves . to determine the dependence of the density of states on @xmath0 we have investigated four distinct temperatures , @xmath14 , @xmath15 , @xmath16 and @xmath17 k. previous numerical estimates based on the eos indicate @xmath18 k @xcite .
to study the phase behavior of the system we implement successive umbrella sampling ( sus ) mc simulations @xcite , from which we evaluate the probability density @xmath19 for the values of @xmath2 sampled by the equilibrium system at fixed @xmath0 , @xmath1 , and volume @xmath20 , and in which the number of molecules @xmath21 in the system fluctuates . in the sus method , the pertinent range of @xmath2 to be investigated , written in terms of the lower and upper number of molecules ( respectively @xmath22 and @xmath23 ) , is divided into many small overlapping windows of size @xmath24 . for each window , a separate grand canonical mc simulation monitors how often a state of @xmath21 particles is visited .
moreover , the simulations are constrained using appropriate boundary conditions on @xmath21 , such that deletions or insertions that would cause @xmath21 to vary outside the range assigned to that window are rejected @xcite .
the density histograms for each window can then be combined to obtain the full @xmath19 curve , by imposing the equality of the probability at the overlapping boundary . in our study @xmath25 ( corresponding to a minimum density @xmath26 g/@xmath27 ) and @xmath28 ( corresponding to @xmath29 g/@xmath27 ) .
we have chosen @xmath30 , i.e. @xmath21 is only permitted to take on one of two adjacent integer values within each window .
the sus method has a number of advantages .
the use of narrow windows in @xmath21 allows an effective sampling of the microstates without the use of biasing functions .
since the windows are independent , all the simulations can be run in parallel , with a gain in throughput that scales linearly with the number of processors employed . in our case ,
approximately 130 processors are used for the calculations at each @xmath0 , one for each window . at the lowest @xmath0 ,
more than two months of simulation time for each window is required for good sampling .
once we obtain @xmath19 at fixed @xmath0 and @xmath1 , histogram reweighting techniques @xcite can be applied to obtain @xmath19 at different values of @xmath1 . keeping track of the coupled density - energy histogram during the sus simulations
also allows us to estimate @xmath19 at different @xmath0 via temperature reweighting .
finally , each window provides accurate information on a specific density , allowing us to compare the results with eos data from previous simulations in the canonical ensemble . to facilitate comparison of our data with future studies , in the following
we will report the activity as @xmath31 ( in units of @xmath32 ) , where @xmath1 is the chemical potential , @xmath33 is @xmath34 , @xmath35 is the boltzmann constant , @xmath36 , @xmath37 is the principal moment of inertia in the direction @xmath38 , @xmath39 is the planck constant and @xmath40 is the de broglie wavelength .
@xmath41 is the quantity that enters in the mc acceptance probability in the insertion - deletion moves .
before discussing the behavior of @xmath19 , we compare our results for the liquid eos as obtained from our sus simulations , with the best available published data . in particular , we focus on the potential energy @xmath42 and the pressure @xmath3 . the pressure is evaluated from the virial using configurations sampled at each density .
[ fig : comparison ] shows results from ref .
@xcite , obtained from molecular dynamics ( md ) simulations , compared with our mc grand - canonical sus results . at all @xmath0 and @xmath2 , we find excellent agreement between these two completely independent numerical methods .
it is worth noting the prominent minimum observed in @xmath42 near @xmath43 g/@xmath27 , the so - called optimal network density @xcite . at this density
the system has the possibility to be able to satisfy all possible bonds , reaching at low @xmath0 the ideal random tetrahedral network state . at lower densities ,
gas - liquid phase separation intervenes , while at larger densities closer packing prevents the system from satisfying the angular and distance constraints required to form linear hydrogen bonds between all pairs of molecules .
in order to establish that the ll phase transition is a genuine liquid - state phenomenon , we must confirm ( i ) that all our simulations are carried out over a time scale that is much longer than the structural relaxation time of the liquid ; and ( ii ) that the time scale for crystal nucleation is much longer than the time scale for liquid - state relaxation .
the separation of these two time scales provides the `` window '' within which the equilibrium behavior of a supercooled liquid can be defined and quantified .
the ability of modern computer simulations to establish liquid - state equilibrium in simulations near a ll critical point is well documented @xcite .
in particular , ref .
@xcite presents the dynamical behavior of the same st2 model as is studied here , as determined from md simulations . in fig .
2(a ) of ref .
@xcite , it is shown that the self diffusion constant @xmath44 for oxygen atoms is greater than @xmath45 @xmath46/s at @xmath47 k for all densities from @xmath48 to @xmath49 g/@xmath27 .
this density range spans the same range within which we find bimodal behavior for @xmath19 at @xmath47 k ( see below ) . for the st2 system
, @xmath50 @xmath46/s corresponds to a range of @xmath38-relaxation times @xmath51 ns @xcite .
time scales well in excess of @xmath52 ns are readily accessible in current md simulations , especially for a small system of a few hundred molecules , as is studied here .
correspondingly , mc simulations of the kind reported here can also easily be run for the number of mc steps required to access equilibrium liquid properties . to demonstrate this , we show in fig .
[ fig : eq ] the @xmath19 histograms generated in our simulations as a function of the computing time invested , for @xmath53 k at a fixed value of the activity @xmath41 .
we find that @xmath19 converges to well - defined values for all densities considered here .
in addition we also check that the energy autocorrelation function decays to zero , in all sus windows , in a number of steps smaller than the simulation length .
we also note that ref .
@xcite demonstrates that the dynamical properties of the liquid state of st2 in the low density region near the optimal network density ( @xmath54 g/@xmath27 ) have a very simple and well - defined dependence on @xmath0 , including in the region where @xmath0 approaches @xmath4 and the ldl emerges as a distinct phase .
to to to
having shown in the previous section that liquid equilibrium is established in our simulations , we next test if crystal nucleation is occurring on the mc time scale of our runs .
this is particularly relevant in view of the conclusions presented in ref .
@xcite , in which it is argued that the ldl phase does not exist and that the only well - defined state of the system at low density is the crystal .
we therefore must quantify the degree of crystalline order in our system at low @xmath0 , especially in the ldl region , where crystallization might occur on a shorter time scale due to the similarity between ice and the liquid in terms of density and local structure .
to to to the degree of crystalline order can be quantified using the steinhardt bond order parameters @xcite based on spherical harmonics of order @xmath55 and @xmath56 , which are particularly suited for discriminating disordered fluid configurations from the open structure of hexagonal ( as well as cubic ) ice . for each particle
we define the complex vector , @xmath57 where the sum is over the @xmath58 neighbors of particle @xmath59 .
@xmath60 is a spherical harmonic of order @xmath61 and @xmath62 , and @xmath63 is a unit vector pointing from the oxygen atom on molecule @xmath59 to that on molecule @xmath64 . in the case of @xmath55 two molecules
are considered neighbors if their oxygen - oxygen distance is smaller than @xmath65 nm , the position of the first minimum of the radial distribution function . in the case of @xmath56 , to be consistent with ref .
@xcite , we assume @xmath66 and neighbors are defined as the four closest particles . the dot product , @xmath67 where , @xmath68 and @xmath69 is its complex conjugate , determines the degree of orientational correlation between neighboring particles @xmath59 and @xmath64 .
[ fig : cij ] shows the distributions of @xmath70 for @xmath55 and @xmath56 , for several densities at @xmath47 k. the distributions have a single broad peak at higher @xmath2 , and become more bimodal at lower @xmath2 with a peak forming near @xmath71 .
however , the distribution goes to zero at @xmath71 .
the bimodal shape of the @xmath72 distribution at low @xmath2 is characteristic of liquids with well formed tetrahedral networks @xcite .
the figure also shows the same distribution evaluated in a hexagonal ice configuration at the same @xmath0 .
the crystal is characterized by a large peak centered at @xmath73 and a smaller peak located approximately at @xmath74 , with @xmath75 the area of the large peak @xcite . in cubic ice ,
only the peak at @xmath73 is present .
the distribution of @xmath76 is rather density independent , but again very different from the crystalline one . ref .
@xcite specifically investigated the global metric , @xmath77 and the related @xmath62-independent rotational invariant , @xmath78 fig .
[ fig : cij](c ) shows @xmath79 at @xmath47 k evaluated from the different windows of densities as well as for hexagonal ice .
again , the configurations sampled in our sus simulations do not show any sign of crystalline order .
the value of @xmath79 in the liquid state , expected to be zero in the thermodynamic limit , is always small ( @xmath80 ) and is one order of magnitude smaller than the crystal value .
obtained from sus mc simulations at activity values of @xmath81 for @xmath47 k , @xmath82 for @xmath53 k , @xmath83 for @xmath84 and @xmath85 at 260 k ( all in nm@xmath86 ) .
the corresponding values of @xmath3 are respectively @xmath87 , @xmath88 , @xmath89 and @xmath90 mpa . below @xmath84
k , bimodality of the distribution emerges , signaling the appearance of two liquid phases with distinct densities .
( b ) isothermal compressibility as a function of density along isotherms , obtained using eq .
[ eq : compress ] .
the location of the peaks matches the data presented in ref .
, title="fig:",width=283 ] obtained from sus mc simulations at activity values of @xmath81 for @xmath47 k , @xmath82 for @xmath53 k , @xmath83 for @xmath84 and @xmath85 at 260 k ( all in nm@xmath86 ) .
the corresponding values of @xmath3 are respectively @xmath87 , @xmath88 , @xmath89 and @xmath90 mpa . below @xmath84
k , bimodality of the distribution emerges , signaling the appearance of two liquid phases with distinct densities .
( b ) isothermal compressibility as a function of density along isotherms , obtained using eq .
[ eq : compress ] .
the location of the peaks matches the data presented in ref .
, title="fig:",width=283 ] to it is possible that a crystallite large enough to affect system properties , but too small to significantly affect a global measure of crystallinity such as @xmath79 , may be present in the system . to test for this
, we estimate the number of crystal - like particles in the system @xmath91 .
we label particles as being crystal - like using two definitions .
for @xmath55 , we define a particle as crystal - like if it has at least three dot products with neighboring particles satisfying @xmath92 .
this criterion has been used in nucleation studies of tetrahedral systems @xcite and in our system the value of -0.87 is rather generous , in that the @xmath72 distribution for the cubic and hexagonal crystals near @xmath93 falls to zero before reaching -0.87 . for @xmath56 , we similarly define a particle to be crystal - like if it has three neighbors with @xmath94 , a low estimate for the value at which the crystal distribution crosses the liquid distributions near @xmath95 .
we plot the average fraction @xmath96 of crystal - like particles using @xmath55 and @xmath56 at @xmath47 k in fig .
[ fig : cij](c ) .
the largest value for @xmath97 suggests that the average crystal cluster size , in our system of about @xmath98 molecules , is not larger than roughly six molecules , again confirming the lack of crystallinity over the entire density range studied . even if we do not resort to average values , the largest number of crystal - like particles that we ever observe in our system , considering all of the configurations we sample , is @xmath99 , found when using @xmath55 .
our observation that crystal - like particles are rare at all densities , including in the range of the ldl phase , demonstrates that the crystal nucleation process occurs on a much longer time scale than that required for liquid - state relaxation .
further , the fact that those small clusters of crystal - like particles that do occasionally form subsequently disappear , shows that a finite and non - trivial nucleation barrier separates the liquid phase from the crystal phase at all densities .
this demonstrates that the liquid phases simulated here are associated with free energy basins that are distinct from that of the crystal phase .
[ fig : dos](a ) shows @xmath19 for all studied temperatures , at the chemical potential for which the density fluctuations are maximal .
the data show the onset of a bimodal distribution below @xmath84 k , consistent with the existence of a ll critical point , and suggests the occurrence of two thermodynamically distinct liquids phases associated with well - separated free energy basins .
the inset shows the same data on a semi - log scale , to highlight the ability of the sus method to provide an accurate estimate of @xmath19 for over 50 orders of magnitude . from the energy - density probability density @xmath100
obtained from the sus simulations , all possible thermodynamic quantities can be calculated , for all values of chemical potential ( limited only by the noise quality of the data ) .
it is interesting to determine the behavior of the isothermal compressibility @xmath101 , which in terms of the fluctuation in the number of particles is , @xmath102 where @xmath103 is the number density and @xmath35 is the boltzmann constant .
[ fig : dos](b ) shows the compressibility at several @xmath0 . as expected , and in agreement with previous estimates for the @xmath101 extrema locus based on md simulations @xcite , the line of @xmath101 maxima moves to smaller densities ( and hence lower pressure ) as @xmath0 increases away from the critical point .
the statistical quality of our estimate for @xmath100 allows us to predict @xmath19 down to approximately @xmath104 k via temperature reweighting .
the reweighted @xmath19 in fig .
[ fig : coexistence ] shows two well - resolved peaks , with a very shallow minimum between them , indicating that the free energy barrier for the system to jump from the ldl to the hdl phase and vice versa , even in a system of only a few hundred particles , is becoming significantly larger than the thermal energy . in the inset to fig .
[ fig : coexistence ] , we plot @xmath105 to highlight this growing free energy barrier .
while the density of the hdl phase progressively increases on cooling , the density of the coexisting ldl remains essentially constant .
this highlights that the thermodynamic stabilization of the ldl phase ( i.e. the establishment of an equilibrium network of tetrahedrally bonded molecules ) requires a very precise density to be achieved .
this strong coupling between the density , local tetrahedral geometry , and free energy is at the heart of the physics of water .
the results reported in this article are consistent with the previous calculations of liu , et al .
@xcite , and show that the evidence for the proposed critical phenomenon in st2 water is independent of the way the long range interactions are modelled .
in addition , our results are also consistent with the possibility that the free energy of this model , projected onto the density , at low @xmath0 is characterized by two basins , both of which correspond to disordered liquid phases .
we also show that even though the location of the proposed ll critical point lies on the extension of the liquid free energy surface that is metastable with respect to crystal formation , the time required for homogeneous nucleation is sufficiently long in our system to allow for local equilibration in phase space . in sum , the evidence presented here continues to point to the existence of a ll phase transition in supercooled st2 water . to further test for behavior consistent with a ll phase transition
, a number of additional questions should be addressed . in particular , in a finite - sized system
, a bimodal density of states can occur in a non - critical system if the correlation length of the order parameter ( here , the density ) exceeds the system size . a finite - size scaling analysis of the density of states , using a range of system sizes ,
would therefore provide an important test for the proposed ll transition in the st2 model .
it would also be valuable to test if the free energy barrier for crystal nucleation is strongly affected by the system size , and by the constrained cubic geometry of the simulation cell , in order to more fully understand the relationship between the time scale for equilibrating the liquid and for crystal formation .
we recommend such studies for future work .
we thank d. chandler and d. limmer for providing us with a preliminary version of ref .
@xcite pertaining to the interpretation of phenomena associated with a critical point in terms of crystallization .
we thank r.k .
bowles and p.g .
debenedetti for useful discussions .
fs acknowledges support from erc-226207-patchycolloids .
is - v thanks the dipartimento di fisica , _ sapienza _ universit di roma , for its hospitality and acknowledges support from acenet and nserc .
php thanks nserc and the crc program . | we perform successive umbrella sampling grand canonical monte carlo computer simulations of the original st2 model of water in the vicinity of the proposed liquid - liquid critical point , at temperatures above and below the critical temperature .
our results support the previous work of y. liu , a.z .
panagiotopoulos and p.g .
debenedetti [ j. chem
. phys . * 131 * , 104508 ( 2009 ) ] , who provided evidence for the existence and location of the critical point for st2 using the ewald method to evaluate the long - range forces .
our results therefore demonstrate the robustness of the evidence for critical behavior with respect to the treatment of the electrostatic interactions .
in addition , we verify that the liquid is equilibrated at all densities on the monte carlo time scale of our simulations , and also that there is no indication of crystal formation during our runs .
these findings demonstrate that the processes of liquid - state relaxation and crystal nucleation are well separated in time .
therefore , the bimodal shape of the density of states , and hence the critical point itself , is a purely liquid - state phenomenon that is distinct from the crystal - liquid transition . |
there is gap in the scientific analysis of the fraction ply methods one of the best methods of search in computer chess and other strategy games . as hans berliner pointed out about the scheme of partial depths ,
... the success of these micros ( micro - processor based programs ) attests to the efficacy of the procedure .
unfortunately , little has been published on this .
this research has the objective of developing a theoretical model of the partial depth scheme based on information theory , implementing it and providing experimental evidence for the method and for the model .
an introduction to games theory and information theory is given in the background section .
a model based on the principles of information theory is outlined and then the formula for partial depths scheme is calculated .
search experiments are performed and then the results are interpreted . in the appendix can be found an introduction to some concepts in chess , and to the axioms of information theory .
an important mathematical branch for modeling chess is games theory , the study of strategic interactions .
assuming the game is described by a tree , a finite game is a game with a finite number of nodes in its game tree .
it has been proven that chess is a finite game .
the rule of draw at three repetitions and the 50 moves rule ensures that chess is a finite game .
sequential games are games where players have some knowledge about earlier actions .
a game is of perfect information if all players know the moves previously made by all players .
zermelo proved that in chess either player @xmath0 has a winning pure strategy , player @xmath1 has a winning pure strategy , or either player can force a draw .
a zero sum game is a game where what one player looses the other wins .
chess is a two - player , zero - sum , perfect information game , a classical model of many strategic interactions .
by convention , w is the white player in chess because it moves first while b is the black player because it moves second .
let m(x ) be the set of moves possible after the path x in the game has been undertaken .
w choses his first move @xmath2 in the set m of moves available .
b chooses his move @xmath3 in the set m(@xmath2 ) : @xmath3 @xmath4 m(@xmath2 ) then w chooses his second move @xmath5 , in the set m(@xmath2,@xmath3 ) : @xmath5 @xmath4 m(@xmath2,@xmath3 ) then b chooses his his second move @xmath6 in the set m(@xmath7,@xmath3,@xmath5 ) : @xmath6 @xmath4 m(@xmath2,@xmath3,@xmath5 ) at the end , w chooses his last move @xmath8 in the set m(@xmath2 , @xmath3 , ... , @xmath9 , @xmath10 ) . in consequence
@xmath8 @xmath4 m(@xmath2 , @xmath3 , ... , @xmath9 , @xmath10 ) let n be a finite integer and m , m(@xmath2 ) , m(@xmath2,@xmath3), ... ,m(@xmath2 , @xmath3 , ... , @xmath9 , @xmath10,@xmath8 )
be any successively defined sets for the moves @xmath2,@xmath3, ... ,@xmath8,@xmath11 satisfying the relations : @xmath12 and @xmath13 a realization of the game is any 2n - tuple ( @xmath2 , @xmath3 , ... , @xmath9 ,
@xmath10,@xmath8,@xmath11 ) satisfying the relations ( 1 ) and ( 2 ) a realization is called variation in the game of chess .
let r be the set of realizations ( variations ) , of the chess game .
consider a partition of r in three sets @xmath14 , @xmath15 and @xmath16 so that for any realization in @xmath14 , player1 ( white in chess ) wins the game , for any realization in @xmath15 , player2 ( black in chess ) wins the game and for any realization in @xmath16 , there is no winner ( it is a draw in chess ) .
then r can be partitioned in 3 subsets so that @xmath17 w has a winning strategy if @xmath18 @xmath2 @xmath4 m , @xmath19 @xmath3 @xmath4 @xmath20 , @xmath18 @xmath5 @xmath4 m(@xmath2,@xmath3 ) , @xmath19 @xmath6 @xmath4 m(@xmath21 , @xmath22 , @xmath23 ) ...
@xmath18 @xmath8 @xmath4 m(@xmath2,@xmath3, ...
,@xmath9,@xmath10 ) , @xmath19 @xmath11 @xmath4 m(@xmath2,@xmath3, ...
,@xmath9,@xmath10,@xmath8 ) , where the variation @xmath24 w has a non - loosing strategy if @xmath18 @xmath2 @xmath4 m , @xmath19 @xmath3 @xmath4 @xmath20 , @xmath18 @xmath5 @xmath4 m(@xmath2,@xmath3 ) , @xmath19 @xmath6 @xmath4 m(@xmath7 , @xmath3 , @xmath5 ) ...
@xmath18 @xmath8 @xmath4 m(@xmath3,@xmath2, ...
,@xmath9,@xmath10 ) , @xmath19 @xmath11 @xmath4 m(@xmath2,@xmath3, ...
,@xmath9,@xmath10,@xmath8 ) , where the variation @xmath25 b has a winning strategy if @xmath18 @xmath3 @xmath4 m , @xmath19 @xmath2 @xmath4 @xmath20 , @xmath18 @xmath6 @xmath4 m(@xmath2,@xmath3,@xmath5 ) , @xmath19 @xmath5 @xmath4 m(@xmath7 , @xmath3 ) ...
@xmath18 @xmath11 @xmath4 m(@xmath2,@xmath3, ...
,@xmath9,@xmath10,@xmath8 ) , @xmath19 @xmath8 @xmath4 m(@xmath2,@xmath3, ...
,@xmath9,@xmath10 ) , where the variation @xmath26 b has a non - loosing strategy if @xmath18 @xmath3 @xmath4 m , @xmath19 @xmath2 @xmath4 @xmath20 ,
@xmath18 @xmath5 @xmath4 m(@xmath2,@xmath3 ) , @xmath19 @xmath5 @xmath4 m(@xmath7 , @xmath3 ) ...
@xmath18 @xmath11 @xmath4 m(@xmath2,@xmath3, ...
,@xmath9,@xmath10,@xmath8 ) , @xmath19 @xmath8 @xmath4 m(@xmath2,@xmath3, ...
,@xmath9,@xmath10 ) ,
where the variation @xmath27 considering a game obeying the conditions stated above , then each of the next three statements are true : ( 1 ) .
w has a winning strategy or b has a non - losing strategy .
b has a winning strategy or w has a non - losing strategy .
+ ( 3 ) . if @xmath16
= @xmath28 , then w has a winning strategy or b has a winning strategy
. + if @xmath16 is @xmath28 , one of the players will win and if @xmath16 is identical with r the outcome of the game will result in a draw at perfect play from both sides .
it is not know yet the outcome of the game of chess at perfect play .
+ the previous theorem proves the existence of winning and non - losing strategies , but gives no method to find these strategies . a method would be to transform the game model into a computational problem and solve it by computational means .
because the state space of the problem is very big , the players will not have in general , full control over the game and often will not know precisely the outcome of the strategies chosen .
the amount of information gained in the search over the state space will be the information used to take the decision .
the quality of the decision must be a function of the information gained as it is the case in economics and as it is expected from intuition . of critical importance in the model
described is the information theory .
it is proper to make a short outline of information theory concepts used in the information theoretic model of strategy games and in particular chess and computer chess .
a discrete random variable @xmath29 is completely defined by the finite set of values it can take s , and the probability distribution @xmath30 .
the value @xmath31 is the probability that the random variable @xmath29 takes the value x. the probability distribution @xmath32 : s @xmath33 [ 0,1 ] is a non - negative function that satisfies the normalization condition @xmath34 the expected value of f(x ) may be defined as @xmath35 this definition of entropy may be seen as a consequence of the axioms of information theory .
it may also be defined independently @xcite . as a place in science and in engineering
, entropy has a very important role .
entropy is a fundamental concept of the mathematical theory of communication , of the foundations of thermodynamics , of quantum physics and quantum computing .
the entropy @xmath36 of a discrete random variable @xmath29 with probability distribution p(x ) may be defined as @xmath37 entropy is a relatively new concept , yet it is already used as the foundation for many scientific fields .
this article creates the foundation for the use of information in computer chess and in computer strategy games in general .
however the concept of entropy must be fundamental to any search process where decisions are taken . some of the properties of entropy used to measure the information content in many systems
are the following : [ [ non - negativity - of - entropy ] ] non - negativity of entropy + + + + + + + + + + + + + + + + + + + + + + + + + + + @xmath38 uncertainty is always equal or greater than 0.if the entropy , h is 0 , the uncertainty is 0 and the random variable x takes a certain value with probability @xmath39 = 1 consider all probability distributions on a set s with m elements .
h is maximum if all events x have the same probability , @xmath40 = @xmath41 if x and y are two independent random variables , then @xmath42 the entropy of a pair of variable x and y is @xmath43 for a pair of random variables one has in general @xmath44 additivity of composite events the average information associated with the choice of an event x is additive , being the sum of the information associated to the choice of subset and the information associated with the choice of the event inside the subset , weighted by the probability of the subset the entropy rate of a sequence @xmath45 = @xmath46 , t @xmath4 n @xmath47 mutual information is a way to measure the correlation of two variables @xmath48 all the equations and definitions presented have a very important role in the model proposed as will be seen later in the article .
@xmath49 @xmath50 if any only if x and y are independent variables .
a necessary condition for a truly selective search given by hans berliner is the following : the search follows the areas with highest information in the tree @xcite `` it must be able to focus the search on the place where the greatest information can be gained toward terminating the search '' .
berliner describes the essential role played by information in chess , however he does not formalize the concept of information in chess as an information theoretic concept . from the perspective of the depth in understanding the decision process in chess the article
@xcite is exceptional but it does not formulate his insight in a mathematical frame .
it contains great chess and computer chess analysis but it does not define the method in mathematical definitions , concepts and equations .
mark winands in @xcite outlines a method based on fractional depth where the fractional ply fp of a move with a category c is given by @xmath51 his approach is experimental and based on data mining as the method presented previously . in the article @xcite david levy , david broughton ,
mark taylor describe the selective extension algorithm .
the method is based on assigning an appropriate additive measure for the interestingness of the terminal node of a path .
consider a path in a search tree consisting of the moves @xmath52 , @xmath53 , @xmath54 and the resulting position being a terminal node .
the probability that a terminal node in that path is in the principal continuation is @xmath55 the measure of the interestingness of a node in this method is @xmath56 + lg [ p ( m_{ij } ] + lg [ p ( m_{ijk } ) ] \label{eq}\ ] ] the problem is to describe the mathematical meaning of information in computer chess , develop the principles and formulas that can be used to control the search and provide experimental evidence for the search heuristic as well as for the role of information gain in obtaining good results at an acceptable cost .
the contributions of this research are the creation of the information theoretical model for search in computer chess , the description of the information gain in computer chess and a scientific explanation of the partial depth scheme .
the paths explored are the areas of the search tree with the highest amount of information gain .
other contributions are , the calculation of information gain for important moves , the calculation of a formula describing the size of the ply added for various moves , the experimental evidence given for the effect of information gain on search for chess problems .
[ [ search - on - informed - game - trees ] ] search on informed game trees + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + in @xcite it is introduced the use of heuristic information in the sense of upper and lower bound but no reference to any information theoretic concept is given .
actually the information theoretic model would consider a distribution not only an interval as in @xcite .
wim pijls and arie de bruin presented a interpretation of heuristic information based on lower and upper estimates for a node and integrated it in alpha beta , proving in the same time the correctness of the method under the following specifications .
consider the specifications of the procedure alpha - beta .
if the input parameters are the following : ( 1 ) n , a node in the game tree , ( 2 ) alpha and beta , two real numbers and ( 3 ) f , a real number , the output parameter , and the conditions : ( 1)pre : alpha @xmath57 beta ( 2)post : alpha @xmath57 f @xmath57 beta @xmath58 , f @xmath59 alpha @xmath60 f(n ) @xmath59 f @xmath59 alpha f @xmath61 beta @xmath60 f(n ) @xmath61 f @xmath61 beta then the procedure alpha - beta ( defined with heuristic information , but not quantified as in information theory ) meets the specification . @xcite considering the representation given by @xcite , assume for some game trees , heuristic information on the minimax value f(n ) is available for any node . the information may be represented as a pair h = ( u , l ) , where u and l map nodes of the tree into real numbers .
u is a heuristic function representing the upper bound on the node .
l is a heuristic function representing the lower bound on the node . for every internal node , n the condition u(n ) @xmath61 f(n ) @xmath61 l(n ) must be satisfied . for any terminal node n the condition u(n )
= f(n ) = l(n ) must be satisfied .
this may even be considered as a condition for a leaf .
a heuristic pair h = ( u , l ) is consistent if + u(c ) @xmath59 u(n ) for every child c of a given max node n and + l(c ) @xmath61 l(n ) for every child c of a given min node n the following theorem published and proven by @xcite relates the information of alpha - beta and the set of nodes visited .
let @xmath62 = ( @xmath63,@xmath64 ) and @xmath65 = ( @xmath66,@xmath67 ) denote heuristic pairs on a tree g , such that @xmath63(n ) @xmath59 @xmath66(n ) and @xmath64(n ) @xmath61 @xmath67(n ) for any node n. let @xmath68 and @xmath69 denote the set of nodes , that are visited during execution of the alpha - beta procedure on g with @xmath62 and @xmath65 respectively , then @xmath68 @xmath70 @xmath69 . in the light of the new description it is possible to reformulate the search problem in a strategy game .
the problem is to plan the search process minimizing the entropy on the value of the starting position considering limits in costs .
the best case is when entropy , or uncertainty in the value of a position becomes 0 with an acceptable cost in search .
this is feasible in chess and it happens every time when a successful combination is executed and results in mate or significant advantage .
it is possible to formulate the problem of search in computer chess and in other games as a problem of entropy minimization .
@xmath71 subject to a limit in the number of position that can be explored .
the entropy of a position can be approximated by the sum of entropy rates of the pieces minus the entropy reduction due the strategical configurations .
this can be expressed as : @xmath72 where @xmath73 represents the entropy of a piece and @xmath74 represents the entropy of a structure with possible strategic importance .
this gives also a more general perspective on the meaning of a game piece .
a game piece can be seen as a stochastic function having the state of the board as entrance and generating possible trajectories and the associated probabilities .
these probabilities form a distribution having an uncertainty associated .
the entropy of a positional pattern , strategic or tactic may be considered a form of joint entropy of the set of variables represented by pieces positions and their trajectory .
the pieces forming a strategic or tactic pattern have correlated trajectories which may be considered as forming a plan .
@xmath75 \label{eq}\ ] ] @xmath76 where @xmath77 is a subset of pieces involved in a strategic pattern and the probabilities @xmath78 represent the probability of realization of such strategic or tactical pattern .
the reduction of entropy caused by strategic and tactical patterns such as double attacks , pins , is determined by both the frequency of such structures and by a significant increase in the probability that one of the sides will win after this position is realized .
we may consider the pieces undertaking a common plan as a form of correlated subsystems with mutual information i(piece1,piece2 , ... ) .
it results that undertaking a plan may result in a decrease in entropy and a decrease in the need to calculate each variation .
it is known from practice that planning decreases the need to calculate each variation and this gives an experimental indication for the practical importance of the concept of entropy as it is defined here in the context of chess .
each of the tactical procedures , pinning , forks , double attack , discovered attack and so on , can be understood formally in this way .
a big reduction in the uncertainty in regard to the outcome of the game occurs , as the odds are often that such a structure will result in a decisive gain for a player .
when such a structure appears as a choice it is likely that a rational player will chose it with high probability .
the entropy of these structures may be calculated with a data mining approach to determine how likely they appear in games .
an approximation if we do not consider the strategic structures would be : @xmath79 [ [ assumption - analysis ] ] assumption analysis : + + + + + + + + + + + + + + + + + + + + the entropy of the position is smaller in general than the sum of the entropies of pieces because there are certain positional patterns such as openings , end - games , various pawn configurations in a chess position which result in a smaller number of combinations , results in order and a smaller entropy .
closer to reality would be such a statement : @xmath80
it is possible to define the information gain during the search process based on the reduction in uncertainty in the following way : @xmath81 where h represents the uncertainty in the value of the position and @xmath82 @xmath83 represents the variation of uncertainty in the current position after a move is made .
it is the information gained after making a move . in the case
when @xmath84 we speak of information gain , if @xmath85 we understand information lost through approximate evaluation or other operation .
it is possible to describe the information gain of the search process by defining the heuristic efficiency @xmath86 when @xmath87 @xmath88 1 the information gain results after a move is @xmath89 this concept may be considered similar to the the concept of information gain for decision trees , the kullback - leibler divergence .
we may see the same principle also here , the higher the difference between entropies , the higher the information gain , which makes very much sense also intuitively and it provides a new theoretical justification for the empirical heuristics of chess and computer chess .
the partial depths method is a generalization of the classic alpha beta in that it offers a greater importance to moves considered significant for the search .
if all moves have the same importance then , the partial depth scheme can be reduced to the ordinary alpha - beta scheme .
it can be described also as an importance sampling search .
the partial depth scheme has been used by various authors .
as hans berliner observed , few has been published about this method @xcite.the contribution of this article goes in this direction .
it is possible to define a function returning the depth : @xmath90 this is a generalization of the classic alpha - beta because in classic alpha - beta @xmath91 depth @xmath92 constant ; if the decision to add a certain depth to the path is dependent only on the current move and position , then if @xmath91 path @xmath93 1 the decision depends only on the current position .
the increase in depth is dependent on the path in this method , where the path is composed of moves @xmath94 , @xmath95 , @xmath96 , .... . in the classic alpha - beta
the depth increase is constant regardless of the type of move
. the principle behind a theory of optimal search should be the allocation of search resources based on the optimality of information gain per cost .
it results that the fraction of a search ply added to the depth of the path with a move should be in inverse proportion to the quality of the move .
the standard approach gives equal importance to all moves , the fraction ply method gives more importance to significant moves .
therefore it must be described a quantitative measure for the quality of a move .
the reduction from the normal depth of 1 ply should be proportional to the quantitative measure of the quality of a move .
the fraction ply fd must decrease with the quality of the move relative to optimal .
the fraction ply added would be equal in this system to the decrease of a full play with the approximate entropy reduction achieved by that move compared to a move having the highest entropy reduction .
for instance for a capture of a rock the entropy reduction is @xmath97 an axiom of efficient search in chess , in computer chess and of efficient search in general should be that the probability of executing a move must be equal to the heuristic efficiency of that move which is equal to the information efficiency of expanding the node resulted after the move .
the same principle can be considered in general for trajectories . by notation ,
let the heuristic efficiency be he and @xmath98 be the probability of a move in category @xmath99 to be executed .
the heuristic efficiency is a fundamental measure of the ability of a search procedure to gain information from the state space .
the heuristic efficiency depends in this analysis on the categories of moves and trajectories defined .
the examples are for moves with individual tactical values , however the analysis can be extended also to tactical plans generated by pins , forks and other tactical patterns . because such analysis would require some readers to look for the meaning of these structures in chess books and also because space considerations the moves generating such configurations would not be presented as examples .
no additional theoretical difficulties would emerge from the introduction of these move categories . the same applies to strategical elements .
following the principles outlined , a formula for the fraction ply can be derived .
@xmath100 considering that @xmath101 and @xmath102 it means @xmath103 for k = 1 , @xmath104 from this , @xmath105 of course a different value than 1 can be given to the constant k and this will propagate without changing the meaning of the equations . the constant k would increase the flexibility of implementations actually , offering more freedom in this direction .
now consider the same equation for the move category with the best information gain .
it means @xmath106 assuming the moves from the best category , the most informational efficient will always be executed in the search , the following condition must be satisfied : @xmath107 then @xmath108 so @xmath109 the cost for execution of any of the two moves is the same . equating this cost
, it results @xmath110 it means @xmath111 which is a very intuitive result . in general , for a @xmath112 , the probability of a trajectory to be explored should be in this system @xmath113 let @xmath114 be the probability that a move is executed and one more ply is added to the search .
the size of the ply added should be function of this probability .
it is logically to consider the size of the play as a quantity increasing with the probability of the move not being executed .
the probability of the move not being executed is @xmath115 therefore assuming an abstraction , a linear relation of the form : size of ply = k * ( probability of a move not being executed ) then the relation between the size of the ply and the probability of the move to be chosen would be for k = 1 @xmath116 this may be considered even a theorem describing the size of the fraction ply in computer chess and even for other exptime problems under the above assumptions and resulting from the above calculations . starting from the previous equation , it is possible to use the relative entropies of pieces and positional patterns to implement the previous formula .
consider the check as the move with the ultimate decrease in entropy because its forceful nature and because it has a higher frequency in the vicinity of the objective , the mate than any other move .
then all the other moves may be rated as function of the check move .
let such value be @xmath117 . here
can be used a constant reflecting the above mentioned properties of such move .
it must be noted that not all checks are equally significant .
several categories of checks can be introduced instead of a single check category .
also in the application , not all checks are equally important , check and capture for example gains a better priority but in this example does not have a smaller depth . as a consequence ,
if the normal increase in search depth is counted as 1 for moves without significance the fractional ply for a check is : @xmath118 then d = 0 in this system because the best move should be always executed and then the depth added should be 0 .
for a capture of queen the entropy rate of the system decreases with @xmath119 .
then the fractional ply for a queen capture is @xmath120 after calculations , d = 0.02 for a capture of rock the entropy rate of the system decreases with @xmath97 .
then the fractional ply for a rock capture is @xmath121 after calculations , d = 1 - 0.776 = 0.223 instead of using the entropy rates for calculating the size of the fractional depth it is possible to use the value of pieces which is strongly correlated for most of the systems with the entropy rate of the pieces.as it can be seen from the calculation above , the higher the differences in entropy between consecutive positions in a variation , the higher the information gain .
this can be understood as a divergence between distributions of consecutive moves .
the more they diverge the higher the information gain after a move . as a test case
it is used a combination which gives us the possibility to define the quality of the response to a position in a precise way .
the meaning of the columns is the following :
+ column 1:experiment number - represents the number of the search experiment + column 2:nodes searched - represents the number of nodes searched in the experiment + column 3:term dividing the reduction in ply - represents the number dividing the term decreasing the size of the normal ply added to the current depth + column 4 : max depth attained - the maximum depth in standard plies attained , here it is added 1 for each ply + column 5 : max uniform depth - the maximum allowed depth in the partial depth scheme considering a step of 6 decreased with a value depending to the quality of the move + column 6 : solved or not - 1 if the case has been solved with the parameters from the other columns + column 7 : step size - the number added to the partial depth for each new level of search in case of moves without importance + the following is the table with the results of the search experiments : [ cols="<,<,<,<,<,<,<,<",options="header " , ] at first a step representing a fraction of 1 has been used .
however , better results have been obtained by using a step bigger than 1 for not so interesting moves .
the cause is the decrease in the sensitivity of the output and of other search dependent parameters in regard to the variations of other parameters and of the positional configurations .
the detection of the variation leading to the objective early decreases the number of nodes searched very much .
the fact that the mate has been found at 13 plies depth after only 20000 nodes searched shows the line to mate has been one of the first lines tried at each level , even without using knowledge .
as it can be seen from the table if the mate is detected relatively fast the number of nodes searched is more than 10 times smaller .
the next plot shows this .
the maximums in the number of nodes represents the configurations ( a set of parameters ) for which the mate has not been fast detected .
the ox represents the number dividing the factor giving importance to some significant moves and on oy it is represented the number of nodes searched .
plot of the increase in number of nodes when the importance given to moves with high information gain is decreased on ox it is represented the virtual depth . on oy it is represented the number of nodes .
as it can be seen , even a deeper search that detects the decisive line will explore less nodes than a shallower search that does not find the decisive line .
for this heuristic and for most of the combinations , when the mate or a strongly dominant line is found fast , the drop in the number of nodes searched is as high as 10 times , even if the uniform search is parametrized for a higher depth . the number of nodes to be searched increases very much with the decrease of importance given to important moves and to lines of high informational value .
the following plot , based on data from the previous table shows the increase in the number of nodes explored with the decrease in the importance given to information gain when the solution is found . the less importance to the information gaining moves and lines is given , the greater the need for a higher amount of nodes to be searched in order to find the solution . on ox it is represented the term dividing the reduction in ply which represents the number dividing the term decreasing the size of the normal ply added to the current depth . on oy it
is represented the number of nodes .
the plot shows the explosion of nodes required to find a solution when the importance given to high information lines is decreased . as the importance given to high information lines
is decreased the number of nodes searched has to be increased .
the importance given to information is decreased so the depth of search must be increased to find the solution .
the following plot has the same significance but for the case when the solution is not found .
the plot of nodes searched vs depth when the solution is detected fast shows a far less pronounced combinatorial explosion then when the solution is not found .
the plot shows the explosion of nodes required to find a solution when the importance given to high information lines is decreased . as the importance given to high information lines
is decreased the number of nodes searched has to be increased .
it increases even faster when the decisive line is not detected . for a high depth of search
, the search cost registers an explosion when no decisive move is found reasonably fast .
when less importance is given to high information gain moves the number of plies has to be increased to compensate this and the number of nodes explodes with the number of plies .
the plot shows the necessary increase of depth when the importance of high information gain moves is decreased . for the case
when the problem is solved the plot is : now we can analyze the data for the cases when the solution is not achieved .
for the case when the position is not solved is a similar plot but the search at the respective depth has been realized at a far greater cost than when the solution has been found fast : the maximum depth achieved decreases with a decrease in the importance given to areas of the tree with high information .
maximum depth vs importance given to information gain . if less importance is given to moves with high information gain more resources are needed for attaining a maximum given depth .
this is the case for solving some combinations .
as it can be seen from the previous plot , the maximal depth has been achieved also when the solution has not been found but as it can be observed from the above table and plots , at an ever increased cost . for the search experiments when
the solution has not been found the highest depth remains the same but this time the cost of resources needed to sustain that depth increased very fast , faster than in the previous plot when the solution has been found .
the search detects the mate even if the maximum length is just one ply deeper than the length of the combination .
even if we keep the maximum depth constant at far greater cost the searches are less likely to find the decisive lines as it can be seen from plots .
as the search has been changed and less importance has been given to interesting moves , the range in the length of the variations became smaller as less energy has been allocated for the most informative search lines than previously and more energy to the less informative lines . after shifting ever more resources from the informative line to other lines , the objective ,
the solution of the combination , has not been attained any more by the best lines who did not have the energy this time to penetrate deep enough .
the best variations did not have any more the critical energy to penetrate the depth of the state space and solve the problem .
the weaker lines were not feasible as a path for finding any acceptable solution .
from this we can understand the fundamental effect of resource allocation . and
how marginal shifts in resources can lead in this context to completely different result .
if somebody used the same depth increase for each move , therefore allocating the resources uniformly to the variations only a supercomputer can go as deep as it is needed for finding the solution to this combination which is not among the deepest . with the introduction of knowledge and heuristics much greater performance would be possible .
the experiment concentrates on one heuristic and its effect on the search is highly significant . in order to solve deep combinations where
some responses are not forced a program must have chess specialized knowledge ( or an extension of the information theoretical model of computer chess to all chess theory ) in order to give importance to variations without active moves but with significant tactical maneuvering between forceful moves such as checks and captures .
[ [ stochastic - modeling - in - computer - chess ] ] stochastic modeling in computer chess + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + in the context of game theory , chess is a deterministic game .
the practical side of decision in chess and computer chess has many probabilistic elements .
the decision is deterministic , but the system that takes the decision is not deterministic , it is a stochastic system .
the human decision - making system and its features such as perception and brain processes are known to be stochastic systems . in the case of computer chess many of the search processes
are also stochastic , as it has been seen from the previous examples .
the implications of the information theoretic model in terms of heuristic development are discussed in this paper .
the extension of the model for more elements of computer chess are left for a different research .
the limitations of the model are given by the ability to detect the information gain resulted from different moves and to quantify the information gain resulted from these moves .
the objective for future research is to explain also other methods from computer chess using the information theoretic model .
applications also in the case of other strategy games are also a future objective .
the model starts from the axiomatic framework of information theory and describes in a formal way the role of information in the efficiency and effectiveness of the heuristics used in computer chess and other strategy board games .
the model proposed considers information in its formal information theoretical meaning as the objective of exploration and the essential factor in the quality of decision in chess and computer chess as well as in other similar games .
the method of partial depths scheme , well known in practice has been described mathematically by observing the fundamental fact that information gain is the criteria that determines the decrease in the uncertainty of the position .
the uncertainty of the position is described in a mathematical way through the concept of entropy .
the information gain describes in a information theoretic way the decrease in uncertainty resulted from making a move . in this way , a quantification of search information is realized .
this refers to entropy as it is understood in information theory but it is possible to build parallels also with thermodynamics .
previous approaches relied on intuitive formulas and descriptions of the best moves in terms of interestingness or in terms of chess theory or using knowledge extracted from the games of strong players .
the approach of the method proposed here is different in that it explains an important method such as the fraction ply method using mathematical methods and formulas that can be derived from the axioms of information theory and determines important coefficients such as the fraction ply associated with moves . the problem of nxn chess is a generalization of the 8x8 chess . it can be expected that the general approach proposed would give a general method for the nxn problem where specialized knowledge is not known and would also provide a method to analyze other exptime - complete problems which can be transformed in the nxn chess .
the method provides a new understanding of chess , a game analyzed scientifically before by scientists such as norbert wiener , john von newumann , allan turing , claude shannon , richard bellman and other famous scientists .
the method proposed generalizes previous approaches and grounds them on information theory a field with a strong theoretical axiomatic system .
it can be expected that the method can provide an example on how to quantify search for difficult problems from classes with high complexity and connect search in computer science also to physics through the common concept of entropy .
[ [ acknowledgment - the - author - acknowledges - with - thanks - his - discussions - with - alberto - giovanni - busetto - and - prof .- j .- buhmann ] ] _ acknowledgment : _ the author acknowledges with thanks his discussions with alberto giovanni busetto and prof .
j. buhmann + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + 100 t.a .
marshland , computer chess methods johnathan schaeffer , the history heuristic and alpha - beta search enhancements in practice claude shannon , programming a computer for playing chess glenn strong , the minimax algorithm t.a .
marshland , m. campbell , parallel search of strongly ordered game trees r.f .
chess the easy way kotov , think like a grandmaster the wikipedia page of games theory as of november 2011 the wikipedia page of computer chess as of november 2011 the wikipedia page of computer go as of november 2011 the wikipedia page of complexity of games as of november 2011 the wikipedia page of thermodynamic entropy as of november 2011 the wikipedia page of information theory as of november 2011 the wikipedia page of thermodynamic entropy as of november 2011 the wikipedia page of thermodynamic entropy and it entropy as of november 2011 alexandru godescu , thesis for the degree of engineer in computer science , pub 2006 alexandru godescu , case study presentation eth zurich 2011 martin j. osborne , ariel rubinstein , a course in game theory theory of games and economic behavior , commemorative edition , ( princeton classic edition ) john von neumann , oskar morgenstern , harold william kuhn and ariel rubinstein s. fraenkel and d. lichtenstein , computing a perfect strategy for n*n chess requires time exponential in n , proc .
automata , languages , and programming , springer lncs 115 ( 1981 ) 278 - 293 and j. comb .
a 31 ( 1981 ) 199 - 214 d.s .
nau , quality of decision versus depth of search on game trees . in : (
2nd ext . ed . ) , ph.d .
thesis , duke university , durham , nc ( 1979 ) .
nau , pathology on game trees : a summary of results . in : proceedings aaai-80
( 1980 ) , pp .
102 - 104 .
nau , an investigation of the causes of pathology in games .
artificial intelligence 19 3 ( 1980 ) , pp .
257 - 278 d.s .
nau , decision quality as a function of search depth in decision trees , j. acm 30 4 ( 1983 ) , pp .
687 - 708 d.s .
nau , pathology on game trees revisited , and an alternative to mini - maxing .
artificial intelligence 21 1 - 2 ( 1983 ) , pp .
687 - 708 m.m .
botvinnik , computers in chess : solving inexact search problems m.m .
botvinnik , computers , chess and long range planning h. j. berliner , a chronology of computer chess and its literature h. j. berliner , the b * tree search algorithm : a best - first proof procedure , artificial intelligence 1978 r. keeny , the chess combination from philidor to karpov , learn tactics from champions b. brgmann , monte carlo go , max - plank - institute of physics h.m .
markovitz ( march 1952 ) `` portfolio selection '' .
the journal of finance 7 ( 1 ) : 77 - 91 t. cover , j. thomas , elements of information theory , second edition marc mezard , andrea montanari , information , physics , and computation wim pijls , arie de bruin , searching informed game trees eliot slater , statistics for the chess computer and the factor of mobility f. etiene de vylder , advanced risk theory , a self contained introduction d e knuth an analysis of alpha - beta pruning , artificial intelligence 1975 g.m .
adelson - velsky , v.l .
arlazarov , m.v .
donskoy , some methods of controlling the tree search in chess programs , artificial intelligence 1975 the wikipedia page of the relative entropy as of november 2011 the wikipedia page of information gain in decision trees as of november 2011 the wikiepdia page of information gain renato renner , lecture notes , quantum information theory , august 16 , 2011 judea pearl , on the nature of pathology in game searching mark winands , enhanced realization probability search david levy , david broughton , mark taylor ( 1989 ) , the sex algorithm in computer chess icga journal , vol .
3 game - tree search algorithm based on realization probability , icga journal , vol . 25 , no . 3 ray solomonoff , the universal distribution and machine learning , the kolmogorov lecture , feb 27 , 2003 , royal holloway , univ .
of london .
the computer journal , vol 46 , no . 6 , 2003 .
jonathan schaeffer , andreas junghanns , search versus knowledge in game - playing programs revisited , aleksander sadikov , ivan bratko , igor konenko , search vs knowledge : empirical study of minimax on krk endgame dana s. nau , mitja lustrek , austin parker , ivan bratko , matjaz gams , when is better not to look ahead ?
mitja lustrek , matjaz gams , ivan bratko , is real - valued minimax pathological ?
mitja lustrek , matjaz gams , ivan bratko , why minimax works : an alternative explanation aleksander sadikov , ivan bratko , igor konenko , bias and pathology in minimax search
.... double minimax(double alfa , double beta , int depth , int k , int type , move mv , double previousval , double virtualdepth ) { move * listnewmoves = ( move * ) new move [ 100 ] ; move mr ; double value = 0 , temp = 0 , ev = 0 ; int c , number ; if ( ( virtualdepth > = maxdepth || depth > = maxextension ) ) { return evaluation(type , mv ) ; } else { if ( tip = = 1 ) { value = -10000 ; } else { value = 10000 ; } generator(mv , listnewmoves , number ) ; for(int i=1 ; i < = number ; i++ ) { listnewmoves[i].eval = fabs ( evaluation(tip , listnewmoves[i ] ) - previousval ) ; double b = -1 ; if ( ischeck ( listnewmoves[i ] ) ) listnewmoves[i].eval + = 10000 ; } if ( number = = 0 ) { if ( tip = = 1 ) if ( !
is_legal_w(mv ) ) return inf_plus ; else return 0 ; } else { if ( !
is_legal_n(mv ) ) return inf_neg ; else return 0 ; } } else for(int k1=1;k1
< = number;k1++ ) { double max = -1 ; int ic = -1 ; for(int c = 1 ; c < = number ; c++ ) { double comp = listnewmoves[c].eval ; if ( comp > max ) { max = listnewmoves[c].eval ; ic = c ; } } mr.eval = listnewmoves[ic].eval ; double evalposition = listnewmoves[ic].eval ; lista_pozitii_urm[ic].eval = -2 ; copy(mr.move , listnewmoves.move ) ; copy ( mr.tabla , listnewmove[ic].tabla ) ; mr.turn = lista_pozitii_urm[ic].turn ; double nextv = evaluation(tip , mr ) ; if ( evalposition > 2000 ) value = - minimax ( -beta , -alfa , depth + 1 , ic , -tip , mr , nextv , virtualdepth ) ; else { double add = log(fabs(0.1 + ( evalposition/100)))/(log(10.0 ) ) + 5.0/log(number + 2 ) ; value = - minimax ( -beta , -alfa , depth + 1 , ic , -tip , mr , nextv , virtualdepth + 6 - add ) ; } if ( value > = alfa ) alfa = value ; if ( alfa > = beta ) { cutoff++ ; break ; } } return alfa ; } } ....
[ [ the - reason - for - presenting - some - concepts - of - chess - theory . ] ] the reason for presenting some concepts of chess theory .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + some of the concepts of chess are useful in understanding the ideas of the paper .
regardless of the level of knowledge and skill in mathematics without a minimal understanding of important concepts in chess it may be difficult to follow the arguments .
it is not essential in what follows vast knowledge of chess or a very high level of chess calculation skills .
however , some understanding of the decision process in human chess , how masters decide for a move is important for understanding the theory of chess and computer chess presented here .
the theory presented here describes also the chess knowledge in a new perspective assuming that decision in human chess is also based on information gained during positional analysis .
an account of the method used by chess grandmasters when deciding for a move is given in a very well regarded chess book .
@xcite .
[ [ combination ] ] combination + + + + + + + + + + + + a combination is in chess a tree of variations , containing only or mostly tactical and forceful moves , at least a sacrifice and resulting in a material or positional advantage or even in check mate and the adversary can not prevent its outcome . the following is the starting position of a combination . the problem is to find the solution , the moves leading to the objective of the game , the mate .
[ [ the - objective - of - the - game . ] ] the objective of the game .
+ + + + + + + + + + + + + + + + + + + + + + + + + + the objective of the game is to achieve a position where the adversary does not have any legal move and his king is under attack .
for example a mate position resulting from the previous positions is : [ [ the - concept - of - variation ] ] the concept of variation + + + + + + + + + + + + + + + + + + + + + + + + + a variation in chess is a string of consecutive moves from the current position .
the problem is to find the variation from the start position to mate . in order to make impossible for the adversary to escape the fate , the mate , it is desirable to find a variation that prevents him from doing so , restricting as much as possible his range of options with the threat of decisive moves .
[ [ forceful - variation ] ] forceful variation + + + + + + + + + + + + + + + + + + + a forceful variation is a variation where each move of one player gives a limited number of legal option or feasible options to the adversary , forcing the adversary to react to an immediate threat .
the solution to the problem , which represents also one of the test cases is the following : \1 .
q - n6 ch ! ; pxq 2 .
bxqnpch ; k - b1 3 .
r - qb7ch ; k - q1 4 .
r - b7 ch ; k - b1 5 .
rxrch ; q - k1 6 .
rxqch ; k - q2 7 .
r - q8 mate [ [ attack - on - a - piece ] ] attack on a piece + + + + + + + + + + + + + + + + + + in chess , an attack on a piece is a move that threatens to capture the attacked piece at the very next move .
for example after the first move , a surprising move the most valuable piece of white is under attack by the blacks pawn .
[ [ the - concept - of - sacrifice - in - chess ] ] the concept of sacrifice in chess + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + a sacrifice in chess represents a capture or a move with a piece , considering that the player who performs the chess sacrifice knows that the piece could be captured at the next turn .
if the player loses a piece without realizing the piece could be lost then it is a blunder , not a sacrifice .
the sacrifice of a piece in chess considers the player is aware the piece may be captured but has a plan that assumes after its realization it would place the initiator in advantage or may even win the game .
for example the reply of the black in the forceful variation shown is to capture the queen . while this is not the only option possible , all other options lead to defeat faster for the defending side .
the solution requires 7 double moves or 13 plies of search in depth .
the entropy as an information theoretic concept may be defined in a precise axiomatic way .
@xcite .
let a sequence of symmetric functions @xmath122 satisfying the following properties : + ( 1 ) normalization : @xmath123 ( 2 ) continuity:@xmath124 is a continuous function of p + ( 3 ) @xmath125 it results @xmath126 must be of the form @xmath127 | the article describes a model of chess based on information theory . a mathematical model of the partial depth scheme
is outlined and a formula for the partial depth added for each ply is calculated from the principles of the model . an implementation of alpha - beta with partial depth is given .
the method is tested using an experimental strategy having as objective to show the effect of allocation of a higher amount of search resources on areas of the search tree with higher information .
the search proceeds in the direction of lines with higher information gain . the effects on search performance of allocating higher search resources on lines with higher information gain
are tested experimentaly and conclusive results are obtained . in order to isolate the effects of the partial depth scheme
no other heuristic is used .
= 1 |
young massive clusters ( ymcs , e.g. , arches , quintuplet , westerlund 1 , rscg , glimpse - co1 : @xcite ) are gravitationally bound stellar systems with masses @xmath110@xmath2m@xmath3and ages @xmath4 100myr @xcite . despite their small numbers and short lifetimes
, they appear to be the ` missing link ' between open clusters and globular clusters ( e.g. , @xcite ) . as such , understanding their formation may provide the link to understanding globular cluster formation .
if ymcs and even globular clusters form in a similar way to open clusters , then our knowledge of nearby stellar systems may potentially be used as a framework to understanding star formation across cosmic time , back to globular cluster formation @xmath0 10 gyr ago .
an understanding of the formation of clusters requires observations of their natal dust and gas well before the onset of star formation . in recent surveys of dense cluster - forming molecular clumps ,
one object , g0.253 + 0.016 , stands out as extreme @xcite .
located at a distance of @xmath08.5kpc , g0.253 + 0.016 lies @xmath0100pc from the galactic centre @xcite . with a dust temperature of @xmath020 k , effective radius of @xmath02.9pc , average volume density of @xmath110@xmath2@xmath5 , mass of @xmath010@xmath6m@xmath3 , @xcite and devoid of prevalent star - formation , g0.253 +
0.016 has exactly the properties expected for the precursor to a high - mass , arches - like cluster . despite being identified over a decade ago , its significance as a precursor to a high - mass cluster has only recently been investigated in detail @xcite .
recent work shows that the volume density threshold predicted by theoretical models of turbulence , which depend on the gas mean volume density and 1-dimensional turbulent mach number , is @xmath010@xmath2@xmath5 for the solar neighbhourhood and is @xmath010@xmath7@xmath5 for the high - pressure inner 200pc environment of the central molecular zone ( cmz ; @xcite ) .
the lack of current star formation within g0.253 + 0.016 is consistent with its location in the cmz and this environmentally dependent volume density threshold for star formation , which is orders of magnitude higher than that derived for the solar neighbourhood @xcite .
because g0.253 + 0.016 is close to virial equilibrium , has sufficient mass to form a ymc through direct collapse , and shows structure on @xmath0 0.3pc scales @xcite , it provides an ideal template to investigate whether or not ymcs form through hierarchical fragmentation , in a scaled up version of open cluster formation . detailed studies of clumps like g0.253 + 0.016 have the potential to reveal the gas structure in protoclusters before star formation has turned on and significantly disrupted its natal material .
if ymcs like the arches are the missing link between globular clusters and more typical open clusters , the detailed study of protoclusters like g0.253 + 0.016 may reveal the initial conditions for star formation not only within high - mass clusters , but potentially for all clusters .
both theoretical studies @xcite and simulations @xcite suggest that high - mass clusters form through the hierarchical merging of overdensities in the interstellar medium .
while some of this hierarchical structure may also be evident in the initial stellar distribution @xcite , it can be easily erased over long time - scales , either via the structure s dispersal ( if unbound ) or relaxation ( if bound ) . in either case , these processes are expected to give rise to spherically symmetric stellar clusters and a background field star population ( e.g. * ? ? ?
* ) . notwithstanding this theoretical picture
, numerical studies of the early phase of cluster evolution often model the gas as a spherically - symmetric background potential ( e.g. * ? ? ?
* ; * ? ? ?
* ) , although notable exceptions exist ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
this disparity between cluster formation theory , which suggests a highly structured initial state , and the early evolutionary modelling , which usually posit a smooth spherical initial state can be resolved by characterising the observed gas structure within high - mass protoclusters before star formation has begun .
the outcome of the high - mass cluster formation process is well - known .
observations show that ymcs are centrally condensed with @xmath8m@xmath3 of stars in their central @xmath9 pc .
thus , the puzzle is then : is this central condensation of stars set by the initial gas structure or do they relax into this small , centrally concentrated region over time ? to answer this question and to distinguish between the models , requires observations of protocluster clumps with a significant improvement in angular resolution , dynamic range , and sensitivity compared to those obtained to date .
fortunately , the new atacama large millimeter / sub - millimeter array ( alma ) provides the necessary improvements and allows us , for the first time , to image protoclusters across the galaxy with comparable physical resolution ( i.e. , @xmath10 0.1pc ) to clouds in the nearby solar neighbourhood .
our goal is to determine the location , mass , and kinematics of the small - scale fragments within g0.253 + 0.016 .
the new alma data reveal the internal structure and kinematics in this high - mass protocluster down to spatial scales of 0.07pc ( fig .
[ three - colour - image ] ) . as such
, these data can be used to determine how the material is distributed within this protocluster .
we imaged the 3 mm ( 90ghz ) dust continuum and molecular line emission across g0.253 + 0.016 using 25 antennas as part of alma s early science cycle 0 . because alma s field of view at this wavelength is @xmath0 1.2 and the cloud is large on the sky ( 3 @xmath11 1 ) ,
a 13-pointing mosaic was obtained to cover its full extent . using alma s band 3 ,
the correlator was configured to use 4 spectral windows in dual polarization mode centred at 87.2 , 89.1 , 99.1 and 101.1ghz each with 1875mhz bandwidth and 488khz ( [email protected] km s@xmath13 ) channel spacing . with
this setup we obtained the 3 mm dust continuum emission and also covered a number of excellent tracers of dense molecular and shocked gas , in particular : hcn , hco@xmath14 , c@xmath15h , sio , so , h@xmath16co@xmath14 , h@xmath16cn , hn@xmath16c , hcc@xmath16cn , hnco , h@xmath15cs ch@xmath17sh , nh@xmath15cho , ch@xmath17cho , nh@xmath15cn , and hc@xmath18n ( for brevity in the text , we refer to the molecular transitions simply by the name of the molecule , e.g. , hcn v=0 j=10 will be called ` hcn ' ; see table [ lines ] for the quantum numbers of the transitions , frequency , critical densities , and excitation energies ) . to build up the signal to noise and improve the @xmath19-coverage , g0.253 + 0.016 was imaged on six different occasions over the period 29 july 1 august 2012 .
we used the alma early science ` extended configuration ' ; the projected baselines range from 13455 m. all data reduction was performed using the casa and miriad software packages .
each individual dataset was independently calibrated ( by our alma support scientist ) before being merged together .
we performed the cleaning and imaging to produce the final data products . because the emission from adjacent 1.7 km s@xmath13 spectral channels are correlated , the final data cubes were resampled to have a velocity width of 3.4 km s@xmath13 per channel .
the alma data have an angular resolution of @xmath01.7 ( @xmath00.07 pc ) , with a 1@xmath20 continuum sensitivity of @xmath0 25@xmath21jy beam@xmath13 and line brightness sensitivity of 1.0 mjy beam@xmath13 per 3.4 km s@xmath13 channel .
because the interferometric alma data are insensitive to emission on the largest angular scales ( @xmath1 1.2 ) , where possible we used single - dish data to fill in the missing zero - spacing information . given that the emission from the clump is extended ( 3@xmath111 ) and the emission from the dust continuum and many of the molecular transitions on spatial scales @xmath1 1.2 significant , it was critical that the zero - spacing information was included to properly sample the @xmath19-plane ( e.g. , @xcite ) .
the combination of the alma data with the zero - spacing information was performed in the fourier domain and was achieved using the corresponding single - dish data as a model during the cleaning process .
the inclusion of the single - dish data in this manner is superior to using alternatives task such as ` feathering ' ( see @xcite for details ) . for this technique to be applicable
there must be an overlapping region of spatial frequencies between the two datasets .
fortunately , for both the continuum and line datasets , existing single - dish data provided the necessary coverage for inclusion with the alma data .
ideally , the single dish data should have the same signal - to - noise as the interferometric data that they are being combined with .
although the signal - to - noise of the single dish data is lower than that of the alma data , it is high enough that the gain from recovering the zero - spacing information outweighs the increased noise that they may introduce . to recover the zero - spacing information for the alma line cubes ( hcn , hco@xmath14 , c@xmath15h , sio , hnco ) , we used data from the malt90 survey which were obtained using the mopra 22 m telescope @xcite . to convert these data from t@xmath22 to the flux @xmath23 in jybeam@xmath13 , we used a scaling factor of 23 k / jybeam@xmath13 , calculated via the equation t@xmath22= 1.36@xmath24 @xmath23 /(@xmath25@xmath26 ) , using a fwhm beam size , @xmath27 , of 38 , wavelength , @xmath28 , of 3 mm , and a main beam efficiency , @xmath25 , of 0.49 @xcite .
for the continuum image , since we do not have a direct measurement of the emission at 3 mm on the larger angular scales , it was necessary to estimate the flux at 3 mm . to achieve this we scaled an image of the dust continuum emission at 500 @xmath21 m ( _ herschel _ , 33 angular resolution ) to what is expected at 3 mm assuming that the flux scales like @xmath29 , where @xmath30 was assumed to be 1.2 .
_ herschel _ data were used for this purpose as these data provide the most reliable recovery of the large scale emission .
we choose to scale the 500 @xmath21 m _ herschel _ image as it is the closest in wavelength to 3 mm , minimising the effect of the assumption of @xmath30 ( see @xcite for more details ) .
figure [ alma - sd ] compares the single dish , alma - only , and combined ( alma and single dish ) images for both the continuum and the hnco integrated intensity . for the 3 mm dust continuum emission , the alma - only image recovers @xmath018% of the total flux .
the images clearly show the significant improvement in angular resolution provided by alma and the recovery of the emission on the large spatial scales provided by the inclusion of the zero - spacing information .
figure [ three - colour - image ] shows a two colour image of g0.253 + 0.016 combining infrared images from the _ spitzer space telescope _
( 24@xmath21 m , @xmath06 angular resolution , shown in cyan ) with the new millimetre image from alma ( 3 mm , @xmath01.7 angular resolution , shown in red ) .
contours show single dish dust continuum emission ( jcmt 450@xmath21 m , @xmath0 7.5 angular resolution ) . identified as an infrared dark cloud ( irdc ) , g0.253 + 0.016 is clearly seen as a prominent extinction feature in mid - ir images . because it is so cold , its thermal dust emission peaks at millimeter / sub - millimeter wavelengths . moreover , since the 3 mm dust continuum emission is optically thin , internal structure is easily revealed by the alma images .
while the overall morphology of the dust continuum emission revealed by alma matches well the mid - ir extinction and the large - scale dust continuum emission provided by the single dish image , the alma image shows significant structure on the small - scale .
the emission is neither smooth nor evenly distributed across the clump ; it is highly fragmented and there is a clear concentration of emission toward its centre .
these small - scale fragments are presumably the cores from which stars may form .
because the 3 mm continuum emission is optically thin , given standard assumptions about the dust emissivity , it can be used to determine the column density and , hence , mass of its small - scale fragments .
recent observations from _ herschel _ of the 170500 @xmath21 m dust continuum emission , show that the dust temperature ( t@xmath31 ) within g0.253 + 0.016 decreases smoothly from 27 k on its outer edges to 19 k in its interior @xcite .
external to g0.253 + 0.016 , the dust temperatures are considerably higher , @xmath135 k. within the cloud and on the spatial scales probed by the _
herschel _ data ( @xmath033 , @xmath01 pc ) , no small volumes of heated dust are apparent .
all of the millimetre emission detected by alma falls within the region for which t@xmath31 was measured to be @xmath4 22k .
thus , we assume that the temperature of the dust detected by alma ranges from 1922k and use a value of 20 @xmath32 1k for all calculations . assuming this dust temperature , a standard value of the gas to dust mass ratio of 100 , a dust absorption coefficient ( @xmath33 ) of 0.27 @xmath34 g@xmath13 ( using @xmath35 = 0.8 @xmath34 g@xmath13 , and @xmath36 @xcite , ) and a @xmath30 of 1.2 @xmath32 0.1 , the intensity of the 3 mm emission can be converted into the h@xmath15 column density , n(h@xmath15 ) , by multiplying the measured fluxes by a single conversion factor of 1.9 @xmath1110@xmath37@xmath38/mjy .
we assume that t@xmath31 is constant across the cloud .
for the small range of measured dust temperatures in g0.253 + 0.016 , this approximation is well justified . given these assumptions , the uncertainties for t@xmath31 and @xmath30
introduce an uncertainty of 10% for the column density , volume density , mass , and virial ratio .
the column densities across g0.253 + 0.016 measured from the alma - only image range from 0.5 - 82 @xmath11 10@xmath39@xmath38 , with a mean value of 1.0 @xmath11 10@xmath39@xmath38 . since this image detects only the highest contrast peaks within the cloud and does not include the large - scale emission , we also determine the column density distribution using the combined ( alma+sd ) image which is more representative of the total column of material associated with g0.253 + 0.016 . when including the emission on all spatial scales , we find that the mean column density is @xmath0 8.6 @xmath11 10@xmath39@xmath38 and an order of magnitude higher compared to molecular clouds in the solar neighbourhood ( see @xcite for a detailed discussion of the column density distribution across g0.253 + 0.016 and its comparison to solar neighbourhood clouds ) . the 3 mm ( 90ghz ) spectrum is rich in molecular lines that span a range in their excitation energies and critical densities ( see table [ lines ] ) .
as such , these observations also provide data cubes from many molecular species and reveal a wealth of information about the gas kinematics and chemistry within g0.253 + 0.016 .
figure [ integrated - intensity - images ] shows , along with the dust continuum image , the integrated intensity images for each of the detected molecular transitions . clearly seen
are the differences in the morphology of the emission from the various tracers , indicating that the chemistry and excitation conditions vary considerably within the clump . to quantify the differences in the morphology of the integrated intensity for each molecular transition compared to the others and the dust continuum emission , we have determined the 2-d cross - correlation function for all combinations of image pairs .
figure [ corr ] shows the derived correlation coefficients for each transition with respect to all others ( the lower half of the figure shows the correlation coefficients via a colour scale , the upper half of the figure shows the derived values ) . for each particular tracer
( labelled along the diagonal ) the molecular transition for which it has the highest correlation coefficient can be found by searching for the darkest square either in the row to its left or in the column below ( the corresponding values are mirrored in the upper portion of the figure ) .
we find that the morphology of the dust continuum emission is not highly correlated with any gas tracer , but is most correlated with the morphology of the emission from the molecules with high excitation energies ( nh@xmath15cho in particular ) .
in addition , we find that the morphology of the emission from hcn and hco@xmath14 are highly correlated with each other and little else , that the shock tracers ( sio and so ) are correlated with each other , and that the dense gas tracers with high excitation energies are well correlated with each other .
figure [ spectra ] shows example spectra from the detected molecular transitions at several locations across the clump ( spectra from a single pixel were extracted at the positions marked on fig .
[ integrated - intensity - images ] and listed in table [ spectra - table ] ) .
the spectra clearly show the complexity of the emission and the differences in intensity and velocity detected from the various tracers .
molecular species with lower excitation temperatures ( e@xmath40/@xmath41 @xmath04k , e.g. , hcn , hco@xmath14 , c@xmath15h ) show emission over a wider range in velocity ( @xmath0 1015 km s@xmath13 ) compared to the isotolopogues ( e.g. , h@xmath16cn , h@xmath16co@xmath14 , hn@xmath16c , and hcc@xmath16cn ; @xmath03 10 km s@xmath13 ) ) and those with higher excitation temperatures ( e@xmath40/@xmath41 @xmath110k , e.g. , ch@xmath17sh , nh@xmath15cho , ch@xmath17cho , h@xmath15cs , nh@xmath15cn , and hc@xmath18n ; @xmath0 3 km s@xmath13 ) .
the shocked gas ( traced by sio and so ) also shows emission over a wide range in velocity towards most locations in the clump ( @xmath0 320 km s@xmath13 ) .
for each molecular transition we characterise the emission via a moment analysis using the definitions of @xcite .
all moments were calculated over the velocity range @xmath1210 km s@xmath13@xmath4v@xmath42@xmath470 km s@xmath13 ( this range is delineated by dotted vertical lines on the spectra ) .
the moment analysis allows us to characterise the emission at every point across the map and to easily compare the emission from the various molecular transitions to each other and to the dust continuum emission .
we compute the zeroth moment ( m@xmath43 , or integrated intensity ) , the first moment ( m@xmath44 , the intensity weighted velocity field ) , and the second moment ( m@xmath15 , the intensity weighted velocity dispersion , @xmath20 ) .
these are defined as @xmath45 where t@xmath22@xmath46 is the measured brightness at a given velocity @xmath47 .
the peak intensity of the emission was also determined via a three - point quadratic fit to each spectrum ( defined as moment @xmath122 ; @xcite ) . note that c@xmath15h has well separated hyperfine components which will affect the first and second moment determinations ( hcn also has hyperfine components , but their separation is less than the observed linewidth for hcn toward g0.253 + 0.016 ) . to include only the highest signal - to - noise emission in the moment analysis , the peak intensity , intensity weighted velocity field , and the intensity weighted velocity dispersion
were only computed at positions across the clump where the emission was above an intensity threshold . for the emission from the molecular species that are the brightest and include the large - scale emission from the single dish data ( i.e.
, hcn , hco@xmath14 , c@xmath15h , sio , and hnco ) we used a threshold in the integrated intensity of @xmath1 20% of the peak . since the noise level is different and the extended emission is included in these cases , this threshold was chosen so that the moments were calculated only when the emission was significantly detected above the noise .
the choice of this threshold was visually checked ( the region over which the moments were calculated are clearly outlined by the edges of the moment maps ) to ensure that it was appropriate for the moment calculation . for the fainter lines with a lower signal to noise that do not include the large - scale emission , we used a threshold in the peak intensity of @xmath1 3.8 mjybeam@xmath13 ( @xmath05@xmath20 ) . at locations within the clump where the emission arises from well separated velocity features within the range @xmath1210
km s@xmath13@xmath4v@xmath42@xmath470 km s@xmath13 , the first moment ( intensity weighted velocity field ) will result in an intensity weighted average of their individual velocities .
in addition , the second moment ( intensity weighted velocity dispersion ) will clearly be larger than the velocity dispersion of the individual velocity components .
thus , while this analysis may not describe well every individual velocity component , it does provide a method by which to easily characterise and compare the emission from the various tracers and the dust continuum across the clump .
figure [ hcn - moments ] shows the integrated intensity image ( zeroth moment , m@xmath43 ) , peak intensity image , intensity weighted velocity field ( first moment , m@xmath44 ) , and also the intensity weighted velocity dispersion ( second moment , m@xmath15 ) derived from the hcn emission toward g0.253 + 0.016 ( all other lines are included on - line in figs .
[ hco+-moments][hc5n - moments ] ) . included on the latter three images are contours of the peak intensity : these highlight the small regions within the clump that are brightest and are included to aid in the comparison to the velocity field and dispersion images .
all images showing the velocity field and dispersions use the same range of values for the colour coding ( i.e. , from @xmath1220 to 60 km s@xmath13 and from 0 to 20 km s@xmath13 respectively ) .
this common colour coding for the different molecular transitions allows easy comparison and readily highlights differences between them .
the images of the velocity field clearly show the well known large - scale velocity gradient across the clump ; most tracers of the dense gas show blue - shifted emission towards more positive declinations and red - shifted emission towards more negative declinations ( i.e. , sio , so , hnco , and h@xmath15cs ) .
some of the more complex molecules ( nh@xmath15cho , ch@xmath17cho , and h@xmath15cs ) are also consistent with this trend .
in contrast to the molecular transitions with lower excitation energies , the transitions from the more complex molecules also reveal emission toward the clump s centre at @xmath020 km s@xmath13 ( coloured green in these images ) . the obvious exception to these trends
is seen in the emission from both hcn and hco@xmath14 , which are clearly red - shifted across the whole clump and with respect to all other tracers .
the trend that these optically thick gas tracers are systematically red - shifted with respect to the optically thin and hot gas tracers , indicates that g0.253 + 0.016 exhibits radial motions ( either expanding or contracting , depending on the relative excitation temperatures of the inner and outer portions of the cloud , see @xcite for an in - depth discussion ) .
the trend of red - shifted hcn and hco@xmath14 emission seen in the single - dish data remains in the interferometric data even when the single dish data are not included .
thus , radial motions are evident on all spatial scales .
figure [ hcn - channels ] shows the channel maps for hcn ( all other channel maps are included on - line in figs .
[ hco+-channels][hc5n - channels ] ) .
the panels show the emission from each spectral channel ( the corresponding velocity is labelled in the top left corner of each panel ) toward positions within the image where the emission was above the previously - defined intensity threshold .
the channel maps also show well the clump s large - scale velocity gradient . on the smaller scales ,
they reveal a complex network of emission features with a complicated velocity structure : there is emission on all spatial scales , the morphology of which ranges from small , compact regions to extended , filamentary structures . toward many of the filamentary features we find homogeneous velocity fields , indicating that they are in fact physically coherent structures .
surprisingly , we find filamentary features in both emission and in absorption .
the absorption filaments are evident in the hcn and hco@xmath14 integrated intensity images and channel maps ( i.e. , figs . [ hcn - moments ] , [ hco+-moments ] and figs . [ hcn - channels ] , [ hco+-channels ] respectively ) .
the most intriguing absorption filaments are those seen in hco@xmath14 that extend for more than 20 km s@xmath13 in velocity . a detailed analysis of these features is discussed in @xcite .
position - velocity diagrams were also generated for each of the detected molecular transitions ( fig . [ pv - diagrams ] ) .
the major axis of the clump is defined along the 0-offset position in right ascension .
the position - velocity diagrams were made by averaging the emission over the clump s minor axis ( i.e. , in right ascension ) and show the emission along the major axis ( declination ) as a function of velocity .
these diagrams also show the complexity in the emission and the systematic red - shift of the hco@xmath14 and hcn emission .
given the complexity of the clump s structure and velocity field , we have employed several statistical methods in order to characterise the observed emission from the dust continuum and the various molecular transitions .
our goal is to answer the questions : how correlated is the emission from the various tracers with respect to one another ? which molecular line , if any , traces well the dust column density ? does turbulence set the observed structure ? what is the structure of the gas ;
is it filamentary and hierarchically distributed or smooth and centrally condensed ?
answers to these questions will reveal both the physical conditions within g0.253 + 0.016 and the mechanisms that are shaping its internal structure .
the combination of molecular transitions covered by these observations trace distinct chemical and physical conditions . for comparison , we group the molecules into four ` families ' : ( 1 ) those that have ` low ' excitation energies e@xmath40/@xmath41 @xmath0 4k ( i.e. , hcn , hco@xmath14 , c@xmath15h ) , ( 2 ) those that are tracers of shocked gas ( i.e. , sio and so ) , ( 3 ) the isotopologues ( i.e. , h@xmath16cn , h@xmath16co@xmath14 , hn@xmath16c , and hcc@xmath16cn ) , and ( 4 ) those that have ` high ' excitation energies e@xmath40/@xmath41 @xmath48 10k ( i.e. , hnco , ch@xmath17sh , nh@xmath15cho , ch@xmath17cho , h@xmath15cs , nh@xmath15cn , and hc@xmath18n ) . figures [ compare_dust_lines0][compare_dust_lines2 ] compare the derived parameters from the moment analysis to the dust column density .
these plots were generated by extracting the values from the moment maps and column density image at every pixel where there were non - zero values in both images . because the number of plotted points within each panel is high ,
the data were binned so that a contour map of the distribution could be made .
contours show the overall distribution and location of the most common values ( the contour levels are shown at 10 to 90% in steps of 10% of the peak number ; the dotted contour outlines the 1% level ) . from top
to bottom the rows show the normalised integrated intensities , normalised peak intensities , intensity weighted velocity field , and intensity weighted velocity dispersion as a function of dust column density . for ease of comparison , the molecular transitions are grouped according to the molecular ` families ' . within each group
the molecular transitions are ordered ( from left to right ) by increasing excitation energies ( see table [ lines ] for the specific values ) .
figure [ compare_dust_lines0 ] shows the plots for hcn and hco@xmath14 . for comparison ,
we show the same plots made including and excluding the large - scale emission ( alma+sd and alma only ; left and right respectively ) to show the effect of its inclusion when calculating the intensity weighted moments . as expected , when including the large - scale emission , we find a clear correspondence between hcn and hco@xmath14 and the lower dust column density material ( i.e. , at column densities @xmath4 5 @xmath11 10@xmath39@xmath38 ) . these molecules clearly show an absence of blue - shifted emission , compared to most others ( compare to figs .
[ compare_dust_lines1 ] and [ compare_dust_lines2 ] ) .
when the large - scale emission is included , the measured intensity weighted velocity dispersion is more centrally peaked ( because the emission is widespread , with a broad line profile , the ` intensity weighting ' will tend to result in a similar calculated value for the intensity weighted velocity despite the emission being detected over a broad range ) .
moreover , a clear trend is evident : the measured velocity dispersions decrease as the dust column density increases .
this trend is not evident in the plots made with the alma - only data because , in this case , there is no emission from the molecular transitions at the same location as the low column density emission , which is required to see the trend .
the shock tracers ( sio and so ) show bright and widespread emission in both position and velocity across g0.253 + 0.016 ( fig .
[ compare_dust_lines1 ] ) .
moreover , their measured velocity dispersions extend to @xmath0 20 km s@xmath13 , which is higher than most other tracers and expected since sio and so trace gas that is stirred up in regions of shocks .
when including the large - scale emission , sio shows the opposite trend in its velocity dispersion compared to hcn and hco@xmath14 : its velocity dispersion increases as the dust column density increases , consistent with increased shocks in the higher density material .
for so , while its typical measured dispersion is low ( @xmath0 2 km s@xmath13 ) , there is a significant tail in this distribution up to dispersions of @xmath0 20 km s@xmath13 .
in contrast to sio , there is no clear trend for an increasing so velocity dispersion with increased dust column density : so shows low velocity dispersions regardless of the column density .
the differences in the velocity dispersions measured for sio and so could be understood in terms of the differences in their excitation energies and critical densities : while so has a higher excitation energy compared to sio , it has a critical density which is an order of magnitude lower .
thus these species may in fact be tracing different material within the clump .
there are many filamentary features found within g0.253 + 0.016 that are detected in sio and so , that have high velocity dispersions , and large steps in velocity over small spatial regions ( i.e. , linear features seen in the moment maps in figs .
[ sio - moments ] and [ so - moments ] ) .
several of these are also evident in the channel maps of h@xmath16cn ( at lsr velocities of @xmath0 1017 km s@xmath13 ; see figs .
[ sio - channels ] , [ so - channels ] , [ h13cn - channels ] ) .
they likely correspond to shock fronts within the clump : their details will be presented in a future paper .
the isotopologues and the molecular transitions with higher excitation energies ( e.g.,h@xmath16cn , h@xmath16co@xmath14 , hn@xmath16c , hcc@xmath16cn , hnco , ch@xmath17sh , nh@xmath15cho , ch@xmath17cho , h@xmath15cs , nh@xmath15cn , and hc@xmath18n ; figs .
[ compare_dust_lines1 ] and [ compare_dust_lines2 ] ) , are detected over a much smaller range in dust column densities and are more sharply peaked at , or slightly above , the mean value of the dust column density ( @xmath0 8.6 @xmath1110@xmath39 @xmath5 , shown as the vertical dotted line ) compared to the molecules with lower excitation energies .
their integrated intensity distribution appears bi - modal , indicating the presence of two distinct integrated intensity values .
their measured velocities span from 0 50 km s@xmath13 ; this velocity distribution is broader compared to that derived from hcn and hco@xmath14 compared to many of the other transitions because they include the large - scale emission from the single - dish data which , due to the ` intensity weighting ' of the moment calculation and in the case of widespread emission with a broad line profile , will tend to result in a similar calculated value for the intensity weighted velocity despite the emission being detected over a broad range .
see figure [ compare_dust_lines0 ] for a comparison of these plots for both hcn and hco@xmath14 including and excluding the single dish data to see this effect . ] .
while the peak intensities for the isotopologues and the molecules with the higher excitation energies are smoothly distributed over a similar range in column densities to their integrated intensities , their velocity dispersion distributions are typically concentrated at low values of the velocity dispersion . with the exception of hnco which is far more abundant than these other molecules , we find that the emission predominantly arises from very narrow velocity features ( @xmath49 2 km s@xmath13 ) .
such narrow velocity dispersions have also been observed toward g0.253 + 0.016 on small spatial scales from other molecular tracers ( i.e. n@xmath15h@xmath14 ( 32 ) ; @xcite ) .
the tail in these distributions to considerably higher velocity dispersions ( dotted contours ) , indicates positions within the clump where the moment analysis has overestimated the velocity dispersion due to multiple velocity components within the spectrum .
this occurs in fewer than 1% of the spectra . the narrow range in dust column density over which the higher excitation energy molecular transitions are detected can be understood in terms of their critical densities .
these molecules are typically detected when the gas has either a high temperature ( t@xmath50@xmath1100k ) or a high density ( n @xmath1 n@xmath51 ) , or both .
because these conditions are needed for both their formation and excitation , the presence of these molecules can be used to pinpoint gas with these physical conditions .
these molecular transitions all have critical densities of @xmath52@xmath5 ( see table [ lines ] ) .
thus , while the gas temperature may be low toward the clump s centre , the volume density may be very high .
if the volume densities are sufficiently higher than their critical densities , the molecules may be excited regardless of the gas temperature .
we see a clear trend that molecules with higher excitation energies and critical densities peak toward the centre of the clump ( compared to those with lower excitation energies and critical densities ) .
the position - velocity diagrams also reveal that the molecular transitions with higher excitation energies and critical densities peak toward the centre of the cloud in both position and velocity ( declination offset of @xmath0 0 , and velocity of @xmath0 20 km s@xmath13 ; e.g. ch@xmath17cho and nh@xmath15cn in fig . [ pv - diagrams ] ) .
these trends are exactly what is expected for a clump with a denser interior .
while the molecular transitions detected by these observations all probe dense gas , they span a range in their critical densities and excitation energies , trace distinct chemical conditions , and often suffer from high optical depths .
as such , they may often trace different material within g0.253 + 0.016 and may not necessarily correlate well with the optically thin dust continuum emission .
one of our key goals is to identify star - forming dust cores within the clump , to measure their sizes , masses , and virial state .
determining these properties will allow us to ascertain the potential of g0.253 + 0.016 to form a cluster and the distribution of masses for its cores . in order to identify bound star - forming cores ,
we need to reliably associate molecular line emission with each dust continuum core in order derive their velocity dispersions .
ideally , we need to identify a molecular transition that is optically thin and is bright enough to be detected in most cores . given the limited velocity resolution of our data , a detailed analysis of the cores virial states are challenging with the current data .
however , we can gauge , overall , which molecular transitions correlate with the dust emission .
we find that emission from the molecular transitions with higher excitation energies typically provide a better morphological match to the dust column density compared to the transitions with lower excitation energies , the isotopologues , or the shock tracers ( compare the morphology of the dust emission to nh@xmath15cho , ch@xmath17cho , hnco and h@xmath15cs in fig .
[ integrated - intensity - images ] and also their 2-d correlation coefficients shown in figure [ corr ] ) .
moreover , since their velocity dispersions are typically narrow ( @xmath0 2 km s@xmath13 ) , these molecules may in fact be tracing individual high column density cores within the clump with very narrow velocity dispersions .
supersonic turbulence is an important mechanism by which overdensities within molecular clouds may be generated @xcite . as such , understanding the degree to which turbulence sets the initial structure within molecular clouds is an important part of understanding core and , ultimately , star formation .
recent work using
the alma dust continuum emission presented here shows that the log - normal shape , dispersion , and mean of its column density probability distribution function ( n - pdf ) very closely matches the predictions of theoretical models of supersonic turbulence in gas given the high - pressure environment in which g0.253 + 0.016 is immersed @xcite .
these results also show that while , overall , the n- pdf shows a log - normal distribution , there is a small deviation from log - normal at the highest column densities , which indicates self - gravitating gas .
this small deviation is coincident with the location of a known water maser @xcite and , thus , pinpoints where star formation is beginning in the cloud . assuming a dust temperature of 20k , this small core ( radius of 0.04pc ) has a mass of @xmath072m@xmath3 and virial ratio , @xmath53 , of @xmath10 1 ( this is an upper limit since the associated molecular line emission is unresolved in velocity , @xmath54 1.4 km s@xmath13 ; see @xcite for details ) . since @xmath53 is close to unity , this core is gravitationally bound and unstable to collapse .
the fact that the n - pdf for g0.253 + 0.016 matches well predictions from turbulent theory suggests that our current understanding of molecular cloud structure derived from the solar neighbourhood also holds in high - pressure environments @xcite .
this has important implications for understanding the initial gas structure for star formation in a range of different environments across the universe @xcite . to further determine the extent to which turbulence shapes the observed gas structure within g0.253 + 0.016 , we have calculated the fractal dimension and the turbulent power spectrum .
these statistical methods are commonly used to describe the structure of molecular clouds in both simulations and observations .
an important characteristic of a protocluster clump , and one that is easily revealed by these data , is the distribution of its material on small spatial scales .
if g0.253 + 0.016 represents a clump that will give rise to a high - mass cluster , then the distribution of the material within it before star formation has commenced can provide important constraints for theories that describe cluster formation . as discussed previously ,
there have been a wide range of theoretical and numerical studies showing that the formation of high - mass stellar clusters should proceed hierarchically , but thus far the degree of hierarchal structure within high - mass protoclusters has not been estimated .
the fractal dimension ( d@xmath55 ) can be used to characterise the structure within a molecular cloud .
it is a simple tool that determines the degree to which the emission is smooth and centrally condensed or structured and hierarchical . since observations of dust or molecular line emission
reveal only the 2 dimensional projection of a 3 dimensional structure , care needs to be taken to accurately convert the measured projected fractal dimension ( d@xmath56 ) to the real three - dimensional fractal dimension ( d@xmath55 ) .
this conversion is often assumed to be d@xmath55 = d@xmath56 + 1 @xcite ; however , using simulations of cloud structure , @xcite derive an empirical relation to convert between d@xmath55 and d@xmath56 ( see their figure 8) . rather than a single conversion factor of unity for all values of d@xmath56 , their simulations suggest that d@xmath55 may actually increase linearly as d@xmath56 decreases .
previous observations derive projected fractal dimensions d@xmath57 1.35 ( see @xcite and references therein ) leading to a quoted d@xmath55 for molecular clouds of @xmath0 2.3 @xcite . to calculate the projected fractal dimension ( d@xmath56 ) for g0.253 + 0.016 , we employ the perimeter - area method @xcite which is often used to measure the fractal dimension of interstellar gas clouds ( e.g. , @xcite ) .
structures were defined as connected regions within a contour at a given intensity level .
the area ( a ) was calculated as the sum of the number of pixels within the contour , the perimeter ( p ) was calculated as the number of pixels externally adjacent to the contour level . by varying the contour levels from the image minimum to its maximum , values for p and a were determined on many spatial scales . with a power - law fit of the form p
@xmath58 a @xmath59 , we can determine the perimeter - area fractal dimension d@xmath56 .
a smooth structure will have d@xmath56 = 1 , while a very convoluted structure will have d@xmath56 = 2 @xcite .
figure [ fractal - dimension ] shows the perimeter - area plots measured from contours of the dust and line integrated intensity emission . for this analysis
we use , where possible , the data that includes the large - scale emission .
since this analysis is based on contours within an image , the effect of including the large - scale emission simply increases the number of contours to lower levels of the intensity which translates to an increase in the number of points plotted at large perimeters and areas .
their inclusion has little effect on the derived slope . in all panels
the solid lines show two extreme values for d@xmath56 ( 1 for smooth , centrally condensed emission , 2 for a highly fractal cloud ) .
the dotted lines show linear fits to the data points : the derived values for d@xmath56 range from 1.41.7 , depending on the molecular tracer . for the dust continuum that includes the large - scale emission
, we find a d@xmath56 of 1.6 ( plotted in the top left panel , marked as alma+sd ) . for comparison we also calculate the fractal dimension derived from only the large - scale dust continuum emission ( in the same panel marked as sd ) , and find d@xmath56 to be @xmath01.2 , consistent with its smooth distribution . since the fractal dimension is best indicated by emission that is optically thin , i.e. , from the dust emission and the molecular transitions with higher excitation energies , we select the derived d@xmath56 from these molecules to characterise the emission from within g0.253 + 0.016 ( i.e. , d@xmath56 @xmath0 1.6 ) . using the empirical conversion from the two - dimensional fractal dimension
d@xmath56 to the three - dimensional fractal dimension d@xmath55 from @xcite , we find that d@xmath55 @xmath0 1.8 for g0.253 + 0.016 .
simulations of compressively driven turbulence in the ism @xcite with mach numbers , @xmath60 , of @xmath0 5.5 , reveal typical values for d@xmath55 of 2.32.6 ( which correspond to d@xmath56 @xmath0 1.31.5 using the empirical relation derived by @xcite ) .
since @xmath60 is @xmath0 30 for g0.253 + 0.016 , it has a much higher level of turbulence compared to most other molecular clouds . given this higher level of turbulence , it may be expected that the sonic length may be lower and the degree of fragmentation might be higher in g0.253 + 0.016 ( i.e. a higher d@xmath56 , and thus , lower d@xmath55 for higher @xmath60 ) . while detailed simulations are necessary to determine the effect of @xmath60 on the fractal dimension , our results are consistent with a marginally higher value for d@xmath56 for g0.253 + 0.016 when compared to typical molecular clouds .
this analysis suggests that the clump is highly fragmented on small spatial scales and its structure consistent with that produced by compressively driven turbulence .
the power spectrum provides a measure of how turbulence is propagated within a molecular cloud and it identifies the preferential spatial scale , if any , at which the structure changes . on large scales
a break in the power spectrum indicates the spatial scale at which turbulent energy is injected , on the small scales it can indicate the spatial scales at which either turbulence is dissipated or injected . measuring the small - scale break point of the power spectrum can therefore reveal the spatial scale at which gravity overcomes the turbulent pressure or the spatial scale at which energy may be injected ( i.e. from protostellar outflows ) .
we use the dust continuum image and several of the integrated intensity images to compute the power spectra .
because the turbulent power spectrum is a statistical measure of the energy on various spatial scales , it is important that the emission is well detected and well sampled across the clump . as such
, we perform this analysis using hcn , hco@xmath14 , hnco , and sio because their data include the zero - spacing information and the emission is detected with high signal - to - noise ratios at all spatial scales across g0.253 + 0.016 .
figure [ power - spectrum ] shows the annular averages of each of the power spectra , p(k ) , as a function of the spatial frequency , k. the power spectra were generated using a fast fourier transform of the images . in all panels the dotted vertical line marks the spatial frequency corresponding to the angular resolution of the data : the power spectra at spatial frequencies greater than this are not well sampled by these data . in cases where the power spectrum is close to a power law ( i.e. , for hcn , hco@xmath14 , hnco , and sio emission ) we fit it with a single power law to derive @xmath61 where , p(k ) @xmath58 k@xmath62 . in these cases , we find that @xmath63 3 . a power - law slope of 3 is predicted for a line intensity power spectrum of an optically thick medium @xcite .
thus , the derived power spectrum slope of @xmath0 3 for these molecular transitions are consistent with them being optically thick toward g0.253 + 0.016 .
in contrast , for the dust continuum emission which is optically thin , the power spectrum shows a clear break . while this break occurs at large spatial frequencies ( i.e. small spatial scales ) , it is lies above the spatial frequencies corresponding to the angular resolution of these data ( marked with the vertical dotted line ) . for this power spectrum
we fit two distinct power laws and derive their indicies and the break point . for small values of k (
i.e. , large spatial scales ) the derived power - law index is @xmath63 2.9 compared to @xmath63 7.5 for large values of k ( i.e. , small spatial scales ) .
values for @xmath61 of @xmath0 2.8 are observed toward other star - forming regions such as perseus , taurus and rosetta @xcite .
the trend of a steeper spectrum on smaller spatial scales implies that the higher column density material is concentrated on smaller spatial scales , while the lower column density material is distributed on larger spatial scales .
this is consistent with the distribution of the emission in the dust continuum image .
since the effect of gravity on small scales will cause a steepening of the power - law slope at large spatial frequencies , the spatial scale at which gravity may dominate can be determined from the break point in the power spectrum . from the dust continuum emission power spectrum , we find that the break occurs at spatial frequencies corresponding to size scales of @xmath00.09pc .
this size scale is comparable to the size of the small region of self - gravitating gas that was identified as the deviation at high column densities via the n - pdf @xcite . as such
, this lends support to the idea that gravity has overcome the internal pressure on these small spatial scales . for cores within molecular clouds to collapse into stars ,
they need to be sufficiently dense that their self - gravity overcomes the thermal pressure .
the jeans length , defined as the minimum length scale for gravitational fragmentation to occur , was determined for g0.253 + 0.016 using the relation @xmath64 where @xmath65 is the sound speed and n(h@xmath15 ) the volume density . while the measured dust temperature toward g0.253 + 0.016 is @xmath0 20k @xcite , recent observations
indicate that the gas temperature is considerably higher , @xmath0 65k @xcite , leading to the speculation that the gas may be heated by the dissipation of turbulent energy and/or cosmic rays rather than by photon heating .
these observations are supported by sph modelling of the dust and gas temperature distribution in g0.253 + 0.016 which suggest that the gas and dust temperatures are not coupled and that the high interstellar radiation field ( isrf ) and cosmic ray ionisation rate ( crir ) in the cmz are responsible for the discrepant dust and gas temperatures @xcite . assuming a global gas temperature for g0.253 + 0.016 of 65k
@xcite the corresponding gas sound speed , @xmath65 , is 0.5 km s@xmath13 . using an average volume density for this material of @xmath0 10@xmath6@xmath5 , we derive @xmath66to be 0.10pc .
this value is comparable to the size scale derived from the break point in the power spectrum of the dust continuum emission and of the size of the self - gravitating core identified via the n - pdf . extrapolating the global size - linewidth relation in figure [ profiles ] ( top right panel ) to smaller scales , suggests that the line - widths are subsonic on spatial scales of @xmath00.1pc ( the sonic length is @xmath0 0.2pc ) . as such ,
since the sonic length exceeds the derived break point in the power spectrum , the thermal jeans length is indeed the relevant quantity for describing the minimum size scale for gravitational fragmentation .
an absence of a similar break in the power spectra derived from the molecular line emission is consistent with the transitions being optically thick and the dust continuum optically thin .
thus , the small , high - column density features detected in the optically thin dust continuum emission may have formed because gravity was able to overcome the thermal pressure on these small scales .
observations show that ymcs like the arches are very centrally condensed , with stellar masses of @xmath110@xmath2m@xmath3 within their central 0.5pc @xcite .
if g0.253 + 0.016 is the molecular cloud progenitor to a cluster like the arches , then its internal structure can reveal the initial conditions for the formation of such high - mass clusters and whether or not the central concentration of stellar densities arises directly from the initial gas structure . to determine its internal structure , we have calculated the mass and volume density profiles of g0.253 + 0.016 as a function of radius ( left panels of fig .
[ profiles ] ) .
we define each region over which to calculate the mass and volume density using contours of column density ; they were calculated on many spatial scales by varying the contour levels from the lowest to the highest values .
an ` effective radius ' was determined for each region by summing the number of pixels contained within that contour level and assuming that this area , a , forms a circle such that @xmath67 .
the mass within each region was then calculated by summing the column density , the volume density is calculated by dividing the mass by the volume of a sphere with a radius of @xmath68 ( left panels of fig . [ profiles ] ) .
using the detected emission from the various molecular species , we have also calculated the average velocity dispersion and virial ratio as a function of radius ( right panels of fig .
[ profiles ] ) .
the virial ratio is @xmath53 , defined as @xmath69 , where @xmath20 is the measured velocity dispersion , @xmath70 the radius , @xmath71 the gravitational constant , and @xmath72 the mass . for these calculations
the ` effective radius ' was determined from the total number of pixels within the integrated intensity image for each of the molecular transitions . within these regions
the total mass , @xmath72 , was determined from the dust column density summed over the same region , while @xmath20 was determined by averaging the intensity weighted velocity dispersions .
the alma observations of the gas structure within g0.253 + 0.016 show that the material is distributed in a complex , filamentary network of structures that are highly fragmented .
its volume density profile ( lower left panel of fig .
[ profiles ] ) shows that the average volume density increases with smaller radii , suggesting that the material is more concentrated toward the clump s centre .
we find that the volume density profile follows very well a power law for radii @xmath4 2pc with an index of @xmath731.2 .
in contrast to what is expected for a bonner - ebert sphere and plummer models that are often used to describe the volume density profiles within starless cores and globular clusters , the volume density profile for g0.253 + 0.016 shows no flattening at small radii . while the total mass of g0.253 + 0.016 is @xmath0 10@xmath6m@xmath3 , its mass profile shows that the majority of the mass originates from the low column density material in its outer edges , at large radii ( upper left panel of fig .
[ profiles ] ) .
we find that the average velocity dispersion decreases on small spatial scales ( upper panel of fig .
[ profiles ] ) , similar to the results of @xcite . moreover , we find that the derived virial ratios are @xmath1 1 for the molecular emission that is widespread ( i.e.have large effective radii ) but @xmath4 1 for the molecular species that are detected only over smaller regions ( lower panel of fig . [ profiles ] ) .
this trend of decreasing virial ratios as the radius decreases , implies that g0.253 + 0.016 is unbound at large radii , but bound at smaller radii .
thus , its outer envelope of lower density material may be unbound and expanding , while the inner denser region is bound and , in the absence of other supporting mechanisms , may be collapsing . the notion that the g0.253 + 0.016 may have a dense and collapsing interior with a low - density surface that is expanding
was posed recently by @xcite when discussing the nature and origin of the hco@xmath14 absorption filaments seen toward g0.253 + 0.016 .
this idea is supported by the observations of the large scale gas motions which show systematically red - shifted optically thick emission compared to the optically thin and hot gas tracers @xcite . in this scenario , the exterior layers of the cloud have a lower excitation temperature compared to the inner layers of the cloud , because the emission is sub - thermally excited ( n@xmath74n@xmath51 ) .
the expanding outer regions , coupled with a collapsing inner region , is consistent with the scenario in which g0.253 + 0.016 was formed recently as a result of a pericentre passage close to sgr a@xmath75 @xcite while a gravitationally bound and collapsing clump is consistent with what is expected in the very early stage of mass assembly within a protocluster , we find that there is no large concentration of mass in the centre of g0.253 + 0.016 . indeed
, its total mass within a radius of 0.5pc is only @xmath0 6 @xmath11 10@xmath76m@xmath3 : this is insufficient to form an arches - like cluster by more than 2 orders of magnitude .
indeed , recent work characterising the structure within other high - mass , dense clumps in the cmz show that the lack of a strongly peaked central mass distribution may be a general trend within these ymc - progenitor clouds @xcite . thus
, if g0.253 + 0.016 is destined to form a high - mass ( @xmath110@xmath2m@xmath3 ) cluster , then the structure of the resulting cluster must evolve from this initial dispersed , hierarchical structure to a smooth , centrally condensed configuration via a dynamical process ( e.g. @xcite ) . as such , any model for cluster formation that posits that stars form directly from an initially smooth , central concentration of gas with a mass similar to that of the final stellar cluster ( e.g. @xcite ) , is ruled out by these observations .
using observations from the new alma telescope we have conducted a statistical analysis of the small - scale structure and velocity within the high - mass protocluster clump g0.253 + 0.016 .
its 3 mm dust continuum and molecular line emission reveal a complex network of emission features that have a complicated velocity structure .
filamentary structures , that are coherent in velocity , are seen in both emission and absorption .
widespread emission is detected from 17 different molecular transitions that probe a range in distinct physical and chemical conditions . comparing their intensity , velocity dispersion , and velocity fields
, we find that the molecular ` families ' trace the same type of material with the clump : transitions that are optically thick have broad velocity dispersions and trace material on the large scales , while transitions that are optically thin have narrow velocity dispersions and trace the densest material in the clump s centre . indeed ,
the dust column density is best traced by molecules with higher excitation energies and critical densities , consistent with a clump that has a denser interior .
the emission is found to be highly fragmented ( d@xmath77 1.8 ) , consistent with structure that is produced by compressively driven turbulence .
we find a clear break in the turbulent power spectra derived from the optically thin dust continuum emission at a spatial scale of @xmath0 0.1pc .
this size scale is comparable to the jeans length corresponding to the observed gas density and temperature , and also to the size of the small core of self - gravitating gas that was identified from the n - pdf .
thus , these structures might correspond to the spatial scale at which gravity has overcome the turbulent pressure .
we speculate that g0.253 + 0.016 is on the verge of forming a cluster from hierarchical , filamentary structures that arise from a highly turbulent medium .
because the initial gas structure is fractal and hierarchical , its structure is quite different from the smooth , compact , very centrally condensed structure observed in high - mass clusters .
consequently , these observations rule out models of cluster formation that begin with such gas distributions .
the implication is that if stars within ymcs form from hierarchical gas structures such as these observed in g0.253 + 0.016 , then their distribution must change to a more centrally condensed configuration over a dynamical time scale .
thus , any theoretical study that aims to accurately characterise and understand the early phases of cluster formation ought to begin with an initial gas structure that is hierarchical rather than one that is smooth and centrally concentrated .
the authors acknowledge their alma contact scientist , dr crystal brogan , for assistance in preparing the observations and performing the initial data calibration .
we thank phil myers , gary fuller , mark krumholz , and eli bressert for valuable discussions .
j.m.r acknowledges funding support via csiro s julius career award that was used to host a week - long team meeting which laid the foundation for this work .
j.m.r , s.n.l , j.m.d.k , j.b . and n.b .
acknowledge the hospitality of the aspen center for physics , which is supported by the national science foundation grant no .
jj acknowledges funding support from nsf grant as 1211844 .
g.g . acknowledges support for conicyt project pfb-06 .
this paper makes use of the following alma data : ads / jao.alma#2011.0.00217.s .
alma is a partnership of eso ( representing its member states ) , nsf ( usa ) and nins ( japan ) , together with nrc ( canada ) and nsc and asiaa ( taiwan ) , in cooperation with the republic of chile .
the joint alma observatory is operated by eso , aui / nrao and naoj . , s. 2002 , in astronomical society of the pacific conference series , vol .
278 , single - dish radio astronomy : techniques and applications , ed .
s. stanimirovic , d. altschuler , p. goldsmith , & c. salter , 375396 combined infrared and millimetre image of g0.253 + 0.016 ( 24@xmath21 m emission from the _ spitzer space telescope _ shown as cyan traces the heated dust from the galaxy ; 3 mm emission from alma shown as red traces the dust from the clump s interior ) overlaid with contours of the single - dish jcmt 450@xmath21 m emission ( contour levels are 30 to 90% of the peak , in steps of 20% ) .
the scale bar marks a size scale of 1pc .
this cloud is so cold and dense it is seen as an extinction feature against the bright ir emission from the galaxy . with the significant improvement in angular resolution and
because alma sees through to the clump s interior , we are now able to pinpoint the dust and gas within it and characterise its internal structure . , scaledwidth=95.0% ] images toward g0.253 + 0.016 showing both the significant improvement in angular resolution provided by alma and the recovery of the emission on the large spatial scales provided by the inclusion of the zero - spacing information .
_ from left to right : _ single dish , alma - only , and combined alma and single dish images .
the upper row compares images of the 3 mm dust continuum emission ( in units of mjybeam@xmath13 ) , while the lower row compares images of the hnco integrated intensity emission ( in units of mjybeam@xmath13 km s@xmath13 ) . in all images the beam size
is shown in the lower left corner .
these ( and all subsequent ) images are shown in equatorial coordinates : the ( 0,0 ) offset position in r.a .
and dec is 17:46:09.59 , @xmath1228:42:34.2 j2000.,title="fig:",scaledwidth=32.0% ] images toward g0.253 + 0.016 showing both the significant improvement in angular resolution provided by alma and the recovery of the emission on the large spatial scales provided by the inclusion of the zero - spacing information .
_ from left to right : _ single dish , alma - only , and combined alma and single dish images .
the upper row compares images of the 3 mm dust continuum emission ( in units of mjybeam@xmath13 ) , while the lower row compares images of the hnco integrated intensity emission ( in units of mjybeam@xmath13 km s@xmath13 ) . in all images the beam size
is shown in the lower left corner .
these ( and all subsequent ) images are shown in equatorial coordinates : the ( 0,0 ) offset position in r.a .
and dec is 17:46:09.59 , @xmath1228:42:34.2 j2000.,title="fig:",scaledwidth=32.0% ] images toward g0.253 + 0.016 showing both the significant improvement in angular resolution provided by alma and the recovery of the emission on the large spatial scales provided by the inclusion of the zero - spacing information . _ from left to right : _ single dish , alma - only , and combined alma and single dish images .
the upper row compares images of the 3 mm dust continuum emission ( in units of mjybeam@xmath13 ) , while the lower row compares images of the hnco integrated intensity emission ( in units of mjybeam@xmath13 km s@xmath13 ) . in all images
the beam size is shown in the lower left corner .
these ( and all subsequent ) images are shown in equatorial coordinates : the ( 0,0 ) offset position in r.a .
and dec is 17:46:09.59 , @xmath1228:42:34.2 j2000.,title="fig:",scaledwidth=32.0% ] + images toward g0.253 + 0.016 showing both the significant improvement in angular resolution provided by alma and the recovery of the emission on the large spatial scales provided by the inclusion of the zero - spacing information .
_ from left to right : _ single dish , alma - only , and combined alma and single dish images .
the upper row compares images of the 3 mm dust continuum emission ( in units of mjybeam@xmath13 ) , while the lower row compares images of the hnco integrated intensity emission ( in units of mjybeam@xmath13 km s@xmath13 ) . in all images
the beam size is shown in the lower left corner .
these ( and all subsequent ) images are shown in equatorial coordinates : the ( 0,0 ) offset position in r.a .
and dec is 17:46:09.59 , @xmath1228:42:34.2 j2000.,title="fig:",scaledwidth=32.0% ] images toward g0.253 + 0.016 showing both the significant improvement in angular resolution provided by alma and the recovery of the emission on the large spatial scales provided by the inclusion of the zero - spacing information . _ from left to right : _ single dish , alma - only , and combined alma and single dish images .
the upper row compares images of the 3 mm dust continuum emission ( in units of mjybeam@xmath13 ) , while the lower row compares images of the hnco integrated intensity emission ( in units of mjybeam@xmath13 km s@xmath13 ) . in all images
the beam size is shown in the lower left corner .
these ( and all subsequent ) images are shown in equatorial coordinates : the ( 0,0 ) offset position in r.a .
and dec is 17:46:09.59 , @xmath1228:42:34.2 j2000.,title="fig:",scaledwidth=32.0% ] images toward g0.253 + 0.016 showing both the significant improvement in angular resolution provided by alma and the recovery of the emission on the large spatial scales provided by the inclusion of the zero - spacing information . _ from left to right : _ single dish , alma - only , and combined alma and single dish images .
the upper row compares images of the 3 mm dust continuum emission ( in units of mjybeam@xmath13 ) , while the lower row compares images of the hnco integrated intensity emission ( in units of mjybeam@xmath13 km s@xmath13 ) .
in all images the beam size is shown in the lower left corner .
these ( and all subsequent ) images are shown in equatorial coordinates : the ( 0,0 ) offset position in r.a .
and dec is 17:46:09.59 , @xmath1228:42:34.2 j2000.,title="fig:",scaledwidth=32.0% ] + 2-d cross - correlation coefficients for each tracer with respect to all others ( the lower half of the figure shows the correlation coefficients via a colour scale , the upper half of the figure shows the derived values ) . for each particular tracer
( labelled along the diagonal ) the molecular transition for which it has the highest correlation coefficient can be found by searching for the darkest square either in the row to its left or in the column below ( the corresponding values are mirrored in the upper portion of the figure ) .
overall , the dust continuum emission is not highly correlated with any gas tracer , but is most correlated with the morphology of the emission from the molecules with high excitation energies ( nh@xmath15cho in particular ) .
as expected , the tracers within each molecular ` family ' are most correlated with each other.,scaledwidth=99.0% ] moment maps for hcn .
_ upper left : _ integrated intensity map ( units of mjybeam@xmath13 km s@xmath13 ) . _
upper right : _ peak intensity map ( units of mjybeam@xmath13 ) .
_ lower left : _ intensity weighted velocity field ( units of km s@xmath13 ) . _
lower right : _ intensity weighted dispersion ( units of km s@xmath13 ) . overlaid on the latter three maps are contours of the peak intensity ( levels are 30 , 50 , 70 and 90% of the peak ) . in all images the angular resolution is shown in the lower left corner.,scaledwidth=95.0% ] channel maps for hcn .
each panel shows the emission ( in units of mjybeam@xmath13 ) at the displayed velocity ( in units of km s@xmath13 ) .
the angular resolution is shown in the lower left corner of each panel.,scaledwidth=95.0% ] moment maps for hco@xmath14 .
_ upper left : _ integrated intensity map ( units of mjybeam@xmath13 km s@xmath13 ) . _ upper right : _ peak intensity map ( units of mjybeam@xmath13 ) . _
lower left : _ intensity weighted velocity field ( units of km s@xmath13 ) . _
lower right : _ intensity weighted dispersion ( units of km s@xmath13 ) . overlaid on the latter three maps are contours of the peak intensity ( levels are 30 , 50 , 70 and 90% of the peak ) . in all images the angular resolution is shown in the lower left corner.,scaledwidth=95.0% ] moment maps for c@xmath15h .
_ upper left : _ integrated intensity map ( units of mjybeam@xmath13 km s@xmath13 ) . _
upper right : _ peak intensity map ( units of mjybeam@xmath13 ) .
_ lower left : _ intensity weighted velocity field ( units of km s@xmath13 ) . _
lower right : _ intensity weighted dispersion ( units of km s@xmath13 ) . overlaid on the latter three maps are contours of the peak intensity ( levels are 30 , 50 , 70 and 90% of the peak ) . in all images the angular resolution is shown in the lower left corner.,scaledwidth=95.0% ] moment maps for hco .
_ upper left : _ integrated intensity map ( units of mjybeam@xmath13 km s@xmath13 ) . _
upper right : _ peak intensity map ( units of mjybeam@xmath13 ) .
_ lower left : _ intensity weighted velocity field ( units of km s@xmath13 ) .
_ lower right : _ intensity weighted dispersion ( units of km s@xmath13 ) .
overlaid on the latter three maps are contours of the peak intensity ( levels are 70% of the peak ) . in all images
the angular resolution is shown in the lower left corner.,scaledwidth=95.0% ] moment maps for sio .
_ upper left : _ integrated intensity map ( units of mjybeam@xmath13 km s@xmath13 ) . _
upper right : _ peak intensity map ( units of mjybeam@xmath13 ) .
_ lower left : _ intensity weighted velocity field ( units of km s@xmath13 ) . _
lower right : _ intensity weighted dispersion ( units of km s@xmath13 ) . overlaid on the latter three maps are contours of the peak intensity ( levels are 30 , 50 , 70 and 90% of the peak ) . in all images the angular resolution is shown in the lower left corner.,scaledwidth=95.0% ] moment maps for so .
_ upper left : _ integrated intensity map ( units of mjybeam@xmath13 km s@xmath13 ) .
_ upper right : _ peak intensity map ( units of mjybeam@xmath13 ) .
_ lower left : _ intensity weighted velocity field ( units of km s@xmath13 ) . _
lower right : _ intensity weighted dispersion ( units of km s@xmath13 ) . overlaid on the latter three maps are contours of the peak intensity ( levels are 30 , 50 , 70 and 90% of the peak ) . in all images the angular resolution is shown in the lower left corner.,scaledwidth=95.0% ] moment maps for h@xmath16cn .
_ upper left : _ integrated intensity map ( units of mjybeam@xmath13 km s@xmath13 ) . _
upper right : _ peak intensity map ( units of mjybeam@xmath13 ) .
_ lower left : _ intensity weighted velocity field ( units of km s@xmath13 ) . _
lower right : _ intensity weighted dispersion ( units of km s@xmath13 ) . overlaid on the latter three maps are contours of the peak intensity ( levels are 30 , 50 , 70 and 90% of the peak ) . in all images the angular resolution is shown in the lower left corner.,scaledwidth=95.0% ] moment maps for h@xmath16co@xmath14 .
_ upper left : _ integrated intensity map ( units of mjybeam@xmath13 km s@xmath13 ) . _
upper right : _ peak intensity map ( units of mjybeam@xmath13 ) .
_ lower left : _ intensity weighted velocity field ( units of km s@xmath13 ) . _
lower right : _ intensity weighted dispersion ( units of km s@xmath13 ) . overlaid on the latter three maps are contours of the peak intensity ( levels are 50 , 70 and 90% of the peak ) . in all images the angular resolution is shown in the lower left corner.,scaledwidth=95.0% ] moment maps for hn@xmath16c .
_ upper left : _ integrated intensity map ( units of mjybeam@xmath13 km s@xmath13 ) . _
upper right : _ peak intensity map ( units of mjybeam@xmath13 ) .
_ lower left : _ intensity weighted velocity field ( units of km s@xmath13 ) .
_ lower right : _ intensity weighted dispersion ( units of km s@xmath13 ) .
overlaid on the latter three maps are contours of the peak intensity ( levels are 30 , 50 , 70 and 90% of the peak ) . in all images
the angular resolution is shown in the lower left corner.,scaledwidth=95.0% ] moment maps for hcc@xmath16cn .
_ upper left : _ integrated intensity map ( units of mjybeam@xmath13 km s@xmath13 ) . _
upper right : _ peak intensity map ( units of mjybeam@xmath13 ) .
_ lower left : _ intensity weighted velocity field ( units of km s@xmath13 ) . _
lower right : _ intensity weighted dispersion ( units of km s@xmath13 ) . overlaid on the latter three maps are contours of the peak intensity ( levels are 30 and 70% of the peak ) . in all images the angular resolution is shown in the lower left corner.,scaledwidth=95.0% ] moment maps for hnco .
_ upper left : _ integrated intensity map ( units of mjybeam@xmath13 km s@xmath13 ) . _
upper right : _ peak intensity map ( units of mjybeam@xmath13 ) .
_ lower left : _ intensity weighted velocity field ( units of km s@xmath13 ) . _
lower right : _ intensity weighted dispersion ( units of km s@xmath13 ) . overlaid on the latter three maps are contours of the peak intensity ( levels are 30 , 50 , 70 and 90% of the peak ) . in all images the angular resolution is shown in the lower left corner.,scaledwidth=95.0% ] moment maps for ch@xmath17sh .
_ upper left : _ integrated intensity map ( units of mjybeam@xmath13 km s@xmath13 ) . _
upper right : _ peak intensity map ( units of mjybeam@xmath13 ) .
_ lower left : _ intensity weighted velocity field ( units of km s@xmath13 ) . _
lower right : _ intensity weighted dispersion ( units of km s@xmath13 ) . overlaid on the latter three maps are contours of the peak intensity ( levels are 30 and 70% of the peak ) . in all images the angular resolution is shown in the lower left corner.,scaledwidth=95.0% ] moment maps for nh@xmath15cho .
_ upper left : _ integrated intensity map ( units of mjybeam@xmath13 km s@xmath13 ) . _
upper right : _ peak intensity map ( units of mjybeam@xmath13 ) .
_ lower left : _ intensity weighted velocity field ( units of km s@xmath13 ) . _
lower right : _ intensity weighted dispersion ( units of km s@xmath13 ) . overlaid on the latter three maps are contours of the peak intensity ( levels are 50 , 70 and 90% of the peak ) . in all images the angular resolution is shown in the lower left corner.,scaledwidth=95.0% ] moment maps for ch@xmath17cho .
_ upper left : _ integrated intensity map ( units of mjybeam@xmath13 km s@xmath13 ) .
_ upper right : _ peak intensity map ( units of mjybeam@xmath13 )
. _ lower left : _ intensity weighted velocity field ( units of km s@xmath13 ) . _
lower right : _ intensity weighted dispersion ( units of km s@xmath13 ) . overlaid on the latter three maps are contours of the peak intensity ( levels are 30 , 50 , 70 and 90% of the peak ) . in all images the angular resolution is shown in the lower left corner.,scaledwidth=95.0% ] moment maps for h@xmath15cs .
_ upper left : _ integrated intensity map ( units of mjybeam@xmath13 km s@xmath13 ) . _
upper right : _ peak intensity map ( units of mjybeam@xmath13 ) .
_ lower left : _ intensity weighted velocity field ( units of km s@xmath13 ) . _
lower right : _ intensity weighted dispersion ( units of km s@xmath13 ) . overlaid on the latter three maps are contours of the peak intensity ( levels are 30 , 50 , 70 and 90% of the peak ) . in all images the angular resolution is shown in the lower left corner.,scaledwidth=95.0% ] moment maps for nh@xmath15cn .
_ upper left : _ integrated intensity map ( units of mjybeam@xmath13 km s@xmath13 ) .
_ upper right : _ peak intensity map ( units of mjybeam@xmath13 ) .
_ lower left : _ intensity weighted velocity field ( units of km s@xmath13 ) . _
lower right : _ intensity weighted dispersion ( units of km s@xmath13 ) . overlaid on the latter three maps are contours of the peak intensity ( levels are 30 and 70% of the peak ) . in all images the angular resolution is shown in the lower left corner.,scaledwidth=95.0% ] moment maps for hc@xmath18n .
_ upper left : _ integrated intensity map ( units of mjybeam@xmath13 km s@xmath13 ) . _
upper right : _ peak intensity map ( units of mjybeam@xmath13 ) .
_ lower left : _ intensity weighted velocity field ( units of km s@xmath13 ) . _
lower right : _ intensity weighted dispersion ( units of km s@xmath13 ) . overlaid on the latter three maps are contours of the peak intensity ( levels are 50 , 70 and 90% of the peak ) . in all images the angular resolution is shown in the lower left corner.,scaledwidth=95.0% ] channel maps for hco@xmath14 .
each panel shows the emission ( in units of mjybeam@xmath13 ) at the displayed velocity ( in units of km s@xmath13 ) .
the angular resolution is shown in the lower left corner of each panel.,scaledwidth=95.0% ] channel maps for c@xmath15h .
each panel shows the emission ( in units of mjybeam@xmath13 ) at the displayed velocity ( in units of km s@xmath13 ) .
the angular resolution is shown in the lower left corner of each panel.,scaledwidth=95.0% ] channel maps for hco .
each panel shows the emission ( in units of mjybeam@xmath13 ) at the displayed velocity ( in units of km s@xmath13 ) .
the angular resolution is shown in the lower left corner of each panel.,scaledwidth=95.0% ] channel maps for sio .
each panel shows the emission ( in units of mjybeam@xmath13 ) at the displayed velocity ( in units of km s@xmath13 ) .
the angular resolution is shown in the lower left corner of each panel.,scaledwidth=95.0% ] channel maps for so .
each panel shows the emission ( in units of mjybeam@xmath13 ) at the displayed velocity ( in units of km s@xmath13 ) .
the angular resolution is shown in the lower left corner of each panel.,scaledwidth=95.0% ] channel maps for h@xmath16cn .
each panel shows the emission ( in units of mjybeam@xmath13 ) at the displayed velocity ( in units of km s@xmath13 ) .
the angular resolution is shown in the lower left corner of each panel.,scaledwidth=95.0% ] channel maps for h@xmath16co@xmath14 .
each panel shows the emission ( in units of mjybeam@xmath13 ) at the displayed velocity ( in units of km s@xmath13 ) .
the angular resolution is shown in the lower left corner of each panel.,scaledwidth=95.0% ] channel maps for hn@xmath16c .
each panel shows the emission ( in units of mjybeam@xmath13 ) at the displayed velocity ( in units of km s@xmath13 ) .
the angular resolution is shown in the lower left corner of each panel.,scaledwidth=95.0% ] channel maps for hcc@xmath16cn .
each panel shows the emission ( in units of mjybeam@xmath13 ) at the displayed velocity ( in units of km s@xmath13 ) .
the angular resolution is shown in the lower left corner of each panel.,scaledwidth=95.0% ] channel maps for hnco .
each panel shows the emission ( in units of mjybeam@xmath13 ) at the displayed velocity ( in units of km s@xmath13 ) .
the angular resolution is shown in the lower left corner of each panel.,scaledwidth=95.0% ] channel maps for ch@xmath17sh .
each panel shows the emission ( in units of mjybeam@xmath13 ) at the displayed velocity ( in units of km s@xmath13 ) .
the angular resolution is shown in the lower left corner of each panel.,scaledwidth=95.0% ] channel maps for nh@xmath15cho .
each panel shows the emission ( in units of mjybeam@xmath13 ) at the displayed velocity ( in units of km s@xmath13 ) .
the angular resolution is shown in the lower left corner of each panel.,scaledwidth=95.0% ] channel maps for ch@xmath17cho .
each panel shows the emission ( in units of mjybeam@xmath13 ) at the displayed velocity ( in units of km s@xmath13 ) .
the angular resolution is shown in the lower left corner of each panel.,scaledwidth=95.0% ] channel maps for h@xmath15cs .
each panel shows the emission ( in units of mjybeam@xmath13 ) at the displayed velocity ( in units of km s@xmath13 ) .
the angular resolution is shown in the lower left corner of each panel.,scaledwidth=95.0% ] channel maps for nh@xmath15cn .
each panel shows the emission ( in units of mjybeam@xmath13 ) at the displayed velocity ( in units of km s@xmath13 ) .
the angular resolution is shown in the lower left corner of each panel.,scaledwidth=95.0% ] channel maps for hc@xmath18n .
each panel shows the emission ( in units of mjybeam@xmath13 ) at the displayed velocity ( in units of km s@xmath13 ) .
the angular resolution is shown in the lower left corner of each panel.,scaledwidth=95.0% ] | g0.253 + 0.016 is a molecular clump that appears to be on the verge of forming a high mass , arches - like cluster .
here we present new alma observations of its small - scale ( @xmath00.07pc ) 3 mm dust continuum and molecular line emission .
the data reveal a complex network of emission features , the morphology of which ranges from small , compact regions to extended , filamentary structures that are seen in both emission and absorption .
the dust column density is well traced by molecules with higher excitation energies and critical densities , consistent with a clump that has a denser interior .
a statistical analysis supports the idea that turbulence shapes the observed gas structure within g0.253 + 0.016 .
we find a clear break in the turbulent power spectrum derived from the optically thin dust continuum emission at a spatial scale of @xmath00.1pc , which may correspond to the spatial scale at which gravity has overcome the thermal pressure .
we suggest that g0.253 + 0.016 is on the verge of forming a cluster from hierarchical , filamentary structures that arise from a highly turbulent medium . although the stellar distribution within arches - like clusters is compact , centrally condensed and smooth , the observed gas distribution within g0.253 + 0.016 is extended , with no high - mass central concentration , and has a complex , hierarchical structure .
if this clump gives rise to a high - mass cluster and its stars are formed from this initially hierarchical gas structure , then the resulting cluster must evolve into a centrally condensed structure via a dynamical process . |
the nature of galactic formaldehyde ( h@xmath0co ) masers is a growing mystery . while hundreds of galactic oh , h@xmath0o , and ch@xmath6oh masers are known , only five galactic star - forming regions have associated h@xmath0co maser emission . to date , this emission is seen only in the @xmath7 transition at 6 cm wavelength . shortly after the discovery of the first h@xmath0co maser in ngc 7538 ( downes & wilson 1974 ; forster et al . 1980 ) , a radiative pumping model was proposed ( boland and de jong 1981 ) . the h@xmath0co masers discovered subsequently did not meet the conditions required for this mechanism ( gardner et al .
1986 ; mehringer , goss , & palmer 1994 , hereafter mgp94 ; hoffman et al .
2003 ; hereafter h03 ) .
thus , twenty - five years after the discovery of the first h@xmath0co maser , these sources remain rare and the excitation mechanism remains unknown .
sgr b2 , the northernmost component of the extended sgr b radio source , is located within a few hundred pc of the galactic center ( reid et al .
( the distance to the galactic center is assumed to be 8.5 kpc in this paper . )
sgr b2 is comprised of three main star - forming complexes designated north ( n ) , middle or main ( m ) , and south ( s ) , and many smaller hii regions .
the h@xmath0co masers occur throughout sgr b2 , shown in figure [ fig1 ] .
the heating mechanisms and complex chemistry of the region are subjects of ongoing study ( e.g. , gaume & claussen 1990 ; goicoechea et al .
2004 ) .
sgr b2 contains nine individual h@xmath0co maser regions , several of which have multiple velocity components .
all of the masers are unresolved at 1 angular resolution , except for maser c which mgp94 suggest consists of several masers blended within the beam .
these regions are near hii regions distributed over the @xmath83.6 arcmin@xmath9 complex ( mgp94 ) .
whiteoak and gardner ( 1983 ) and mgp94 designated the maser regions with letters ( fig .
[ fig1 ] ) .
the h@xmath0co masers are observed over the velocity range @xmath10 km s@xmath11@xmath12 km s@xmath11 , while other species , such as h@xmath0o masers , are observed over a larger range @xmath13 km s@xmath11@xmath14 km s@xmath11 ( kobayashi et al .
1989 ; mcgrath , de pree , and goss 2004 ) . of the nine maser regions observed in sgr b2 by mgp94
, the g maser was shown to be time variable , at least quadrupling in intensity over 10 yr .
( similarly , h03 found one ngc 7538 feature to triple in intensity over @xmath15 yr . ) as initially noted by whiteoak and gardner ( 1983 ) , all of the sgr b2 h@xmath0co masers lie close to oh , h@xmath0o , ch@xmath6oh , and nh@xmath6 masers .
for most of the masers in mgp94 , the separation to an oh maser was less than 0.05 pc .
recent successes in search techniques for new masers ( araya et al . 2004 , 2005 , 2006a ) and in high - resolution observational techniques ( h03 ) promise to provide empirical constraints for the development of a realistic model for the galactic h@xmath0co maser emission .
the necessary steps in compiling an empirical picture of the h@xmath0co emission in sgr b2 are ( 1 ) detailed imaging of the masers in order to quantify the intrinsic properties of the emission ( e.g. , brightness temperature ) , ( 2 ) assessment of intensity variability in the masers , and ( 3 ) precise astrometry for elucidating spatial relationships between the h@xmath0co masers and more common masers ( oh , h@xmath0o , ch@xmath6oh ) . in this paper
, we present new observations of the h@xmath0co masers in sgr b2 using the the very long baseline array ( vlba ) and very large array ( vla ) of the nrao .
we observed the a and d h@xmath0co masers in sgr b2 using the ten antennas of the vlba and the 27 antennas of the vla as an 11-station vlbi array .
parameters of the observations are summarized in table [ tbl - tab1 ] .
the total observing time was approximately 8.0 hours alternating between the a and d pointing positions which are separated by approximately 1 . at each pointing position
there is a useful correlated field of view of approximately 1 ( e.g. , bridle & schwab 1999 ) . to observe each of the nine h@xmath0co masers in sgr b2 optimally would have required observations at nine pointing centers . because of signal - to - noise considerations , we observed only the two masers measured to be most intense by mgp94 .
the baseline lengths of the vlba+y27 array range from 52 km to 8611 km ; the array is not sensitive to angular scales larger than 025 .
the antennas have right- ( r ) and left- ( l ) circularly polarized feeds from which rr , ll , rl , & lr cross - correlations were formed .
the visibilities were integrated for 8.4 s. the amplitude scale is set using online system temperature monitoring and _ a priori _ antenna gain measurements .
the station delays were determined from observations of j1733 - 130 .
the maser observations were phase referenced to j1745 - 283 ( called w56 in bower et al .
because the properties of this source limit our observations , we discuss it in some detail .
the absolute position uncertainty of this source is 12 milliarcseconds ( hereafter : mas ) ( reid et al . 1999 ) .
bower et al .
find that j1745 - 283 is probably the core of an extragalactic jet source and that its apparent size at 5 ghz ( 30 mas ) is determined by scatter - broadening .
we also observed j1745 - 283 on 22 november 2002 using only the ten vlba stations ( a snapshot observation with a bandwidth of 4 mhz in both polarizations ) .
the 2002 observation yields a deconvolved angular size of @xmath16 mas and an integrated flux density of 28 mjy ; the 2003 observation yields a size of @xmath17 mas and a flux density of 31 mjy .
( the flux density determined by y27 [ the vla alone ] during the 2003 observations was 161 mjy . )
bower et al .
report an angular size @xmath18 mas and peak intensity 89 mjy beam@xmath11 from vlba+y1 observations .
therefore , the flux density must have decreased significantly between 1999 and 2002 .
such variability would not be unusual for an inverted spectrum source like j1745 - 283 ( e.g. , urry & padovani 1995 ) .
our imaging of j1745 - 283 with the vlba makes use of only the inner - most stations ( vla , pie town , los alamos , fort davis , kitt peak , and owens valley ) because the source is resolved on longer baselines . some additional resolution for the maser observations
was gained by self - calibration ; but , because only the shorter baselines in the vlba data were absolutely calibrated in phase , the position registration accuracy of the resulting vlba images is approximately 15 mas . in summary , we could not make use of the full potential resolution of the vlba when using this phase referencing calibrator , and no other suitable nearby source is known .
we detected masers toward the a and d regions .
the f maser is also within the correlated field of view near the a maser , but was not detected with the current sensitivity .
the image and spectrum of the a maser from the vlbi data is shown in figure [ vlba - a ] . because the radio continuum of sgr b2 is fully resolved by the vlba , no continuum subtraction was necessary .
the positions , center velocities ( @xmath19 ) , linewidths ( @xmath20 ) , peak flux densities ( @xmath21 ) , and deconvolved major and minor axes and position angles are summarized in table [ tbl - vlba ] . in parentheses following each entry
are the 1-@xmath22 errors .
the deconvolved sizes of the vlbi images of the a and d masers are @xmath23 mas . at the distance of sgr b2
, this size corresponds to a linear diameter of approximately 80 au , comparable to the sizes ( 30 to 130 au ) observed for the h@xmath0co masers in ngc 7538 and g29.96 - 0.02 using the vlba ( h03 ) .
the brightness temperatures of the sgr b2 a and d masers in the current vlbi data are @xmath24 k and @xmath25 k , respectively , similar to the @xmath26 k observed for other h@xmath0co masers ( h03 ) . no significant linear or circular polarization is detected in either maser ( @xmath27 20% for the strongest maser [ a]).gipsy/ ) and the astronomical image processing system ( aips ) software package ( http://www.nrao.edu/aips/ ) . ] because of the known variability of h@xmath0co masers ( forster et al . 1985 ; h03 )
, the sgr b2 masers were observed in 2005 with the vla in order to access any possible time - variability .
parameters of the observations are summarized in table [ tbl - tab1 ] .
the array configuration available was dnc for an observation period of approximately 7 hours .
two pairs of rr and ll bands were recorded centered at the expected velocities of the a and the d masers .
of the nine maser regions described by mgp94 , eight of the sources were detected .
six velocity components in the regions observed by mgp94 lie outside the velocity range of the current observations .
no new h@xmath0co maser regions were discovered .
the 2005 data have inferior angular resolution ( 10 versus 1 ) but improved spectral resolution ( 0.19 km s@xmath11 versus 1.5 km s@xmath11 ) compared with the 1993 vla observations of mgp94 .
comparison of the current data to the mgp94 data is uncertain due to ( 1 ) the severe blending of the strong h@xmath0co absorption with the nearby masers and ( 2 ) the insufficient velocity resolution of the mgp94 data , which did not spectrally resolve the lines .
variability of the flux density of the masers is apparent and is discussed in
[ disc - vary ] .
the 2005 results are summarized in table [ tbl - vla ] and a spectrum from these data ( region e ) is shown in figure [ vla - e ] .
the c maser region is not discussed in this paper due to confusion of the maser spectra with nearby absorption and continuum emission . with the current 10 angular resolution of the vla data , as with the 1 resolution of mgp94
, none of the masers are spatially resolved . in
[ disc - vary ] , we also compare the current data with the 1983 vla observations of gardner et al .
the velocity resolution of the 1983 data is 0.76 km s@xmath11 and the angular resolution is 125 .
although we resolve many of the maser line profiles in velocity for the first time with the 2005 vla data , most of the linewidths presented in table [ tbl - vla ] agree with the values measured by gardner et al .
( 1986 ) .
no circular polarization is detected in any of the h@xmath0co masers .
this corresponds to a 3-@xmath22 upper limit of approximately 4% circular polarization for the strongest maser ( a ) .
images of radio sources are angularly broadened by scattering by the ionized component of the interstellar medium . in the direction of the galactic center ,
this problem becomes severe ( e.g. , lazio & cordes 1998 ) . as discussed in 2.1 , j1745 - 283 ,
is an extragalactic point - like source whose image is broadened to @xmath830 mas at 6 cm wavelength by scattering ( bower et al .
2001 ) . in this section
we discuss the extent to which this scattering medium affects the current vlbi observations of the h@xmath0co masers in sgr b2 . the observed size for j1745 - 283 , @xmath17 mas , is larger than the observed deconvolved size for the a maser , @xmath28 mas .
this result may be expected even if both sources are intrinsically point - like because the proximity of the maser to the scattering medium results in reduced angular broadening ( e.g. , rickett 1990 ) . nevertheless , as discussed below
, we expect significant angular broadening in the images of the masers . therefore , the deconvolved angular sizes in table [ tbl - vlba ] are upper limits to the intrinsic sizes and the brightness temperatures are lower limits .
gwinn et al . ( 1988 ) quantified the scattered sizes of h@xmath0o and oh masers in sgr b2 , finding a @xmath29 dependence for the broadening . from their determinations of minimum angular sizes of 0.3 mas for 1.35-cm h@xmath0o masers and 100 mas for 18-cm oh masers , we expect a minimum apparent size of approximately 9 mas for the 6.2-cm h@xmath0co masers .
the deconvolved angular sizes of the masers in table [ tbl - vlba ] are in agreement with this expectation . in comparing the current vla observations with earlier vla data , both the differing angular resolution and spectral resolution
must be considered . as discussed in
[ vla - obs ] , both the differences in array configuration and in correlator setup were significant . to compensate for the difference in velocity resolution , in table [ tbl - fd ] we tabulate the velocity - integrated flux density for observations made in 1983 ( gardner et al .
1986 ) , 1993 ( mgp94 ) , and 2003 and 2005 ( this paper ) . however , it is impossible to compensate for emission outside of the velocity range covered in the current observations and for the confusion of features caused by the lower angular resolution .
( the notes in the final column of table [ tbl - vla ] summarize limitations of the current data from these causes . )
the error values in table [ tbl - fd ] are dominated by systematic uncertainties rather than by thermal noise in most cases .
we find that the a maser has approximately doubled in velocity - integrated flux density since 1993 , while b , g , and h increased between 1983 and 1993 , and subsequently decreased .
the velocity - integrated flux densities from regions d , e , f , and i have not changed significantly since 1993 . for the a and d h@xmath0co masers in sgr b2 , a comparison of the current vlba+y27 data with past and current vla data shows two major differences : ( 1 ) only a fraction of the flux density seen with the vla is observed in the vlbi data ( @xmath870% for the a maser ; @xmath840% for the d maser ) ; and ( 2 ) the velocity widths in the vlbi data are more narrow than in the vla data ( @xmath880% for the a maser ; @xmath850% for the d maser ) . in this section we discuss a schematic model for the structure of the emission which addresses both the observed angular distribution and velocity widths .
the flux density detected by the vla but resolved by vlbi baselines must lie at angular scales between @xmath30 mas ( the resolution of the shortest vlba+y27 interferometer spacings ) and @xmath31 mas ( the masers are unresolved in the mgp94 observations with 1 resolution ) , while the flux density observed with the vlbi is emitted by a source @xmath810 mas .
these limits suggest a core - halo source morphology .
a similar morphology was proposed by h03 for the h@xmath0co masers in ngc 7538 and g29.96 - 0.02 .
a model consisting of two coincident circular gaussian components with different angular sizes can reproduce the observed results .
for the a maser , the flux density observed with the vla is about 2000 mjy . to determine the decrease in flux density with increasing projected baseline length
, we use the highest signal - to - noise baselines only : those between y27 and other antennas .
the average flux densities observed on baselines with y27 are : 1400 mjy at 0.71 m@xmath32 ( to pie town , average projected spacing @xmath844 km ) denotes million wavelengths ] ; 1300 mjy at 3 m@xmath32 ; and 1200 mjy at 5 m@xmath32 .
these data are fit by a two component gaussian model with a 600 mjy in a 300 mas halo and 1400 mjy in a 10 mas core .
the brightness temperature of the halo is @xmath33 k ; for the 10 mas core component , @xmath34 k ( table [ tbl - vlba ] , fig .
[ vlba - a ] ) .
a more precise model is justified only after higher signal to noise measurements .
it is reasonable to assume that the background radiation is relatively uniform on 300 mas angular scales because the continuum emission is well resolved on the vlba baselines .
therefore , the difference in brightness temperatures between the core and halo components may be attributed to differences in maser gain .
we estimate that the halo has a gain of approximately 10 ( in the exponential amplification regime ) and the core has a gain of approximately 2000 ( in the saturated regime ) . because the observed angular sizes are upper limits for the intrinsic sizes , the derived brightness temperatures are lower limits for the intrinsic brightness temperatures ; therefore
, the actual gains may exceed the above values .
the gain of the maser medium is dependent upon four factors : the pathlength , the density of h@xmath0co , the level inversion , and the velocity coherence ( see 4.2 of h03 ) .
therefore , one or more of these factors must be significantly different between the core and halo components .
we suggest that the narrower line widths arising from the core component indicate that the velocity coherence is enhanced in the core region compared to the halo , yielding a higher gain .
similar arguments may be applied to the d maser , but the resulting ranges of size and @xmath35 are not well constrained because of the higher noise level on the vla - pie town baseline .
additional maser species in other star - forming regions excited oh ( palmer , goss , & devine 2003 ) , oh in supernova remnants ( hoffman et al.2005 ) , as well as other h@xmath0co masers ( h03 ) exhibit a similar narrowing of line widths between vla and vlba angular scales . h@xmath0co absorption toward sgr b2 has been studied extensively with 10 @xmath36 20 resolution both by martn - pintado et al .
( 1990 ) and mehringer et al .
martn - pintado et al .
( 1990 ) noted that the absorption is dominated by three velocity components ( @xmath19= 55 , 64 , and 80 km s@xmath11 ) , each with @xmath37 .
linewidths of these absorption features ranged from 9 26 km s@xmath11 , i.e. much greater than that of the maser features ( @xmath38 km s@xmath11 ; see table [ tbl - vla ] ) .
mehringer et al .
( 1995 ) provide optical depth profiles toward selected positions .
all but maser e lies within 30 of one of these positions displayed .
( maser e lies at a position with no radio continuum . )
it is striking that the maser velocities , except for maser i ( which is more than a beamwidth away from the position of a displayed profile ) , do not occur in velocity ranges with large h@xmath0co optical depths .
masers a and f occur in a velocity range with @xmath39=1 , but this velocity range is a rather sharp minimum in the optical depth profile at this position .
therefore , we conclude that the gas containing the h@xmath0co masers is distinct from the bulk of the h@xmath0co containing gas observed in absorption .
insight into why the maser containing gas may be distinct is provided by ( 3,3 ) and ( 4,4 ) nh@xmath6 observations made with the vla by martn - pintado et al .
( 1999 ) . with their 3 resolution ,
80 - 90% of the nh@xmath6 emission is resolved out , and only small angular scale features remain . among these features
are a number of rings and arcs , most naturally interpreted as complete or partial shells with linear sizes @xmath82 pc .
the shells are hot ( t@xmath40 50 70 k ) , and the h@xmath0 densities derived for them are typically a factor of 10 greater than those derived from the h@xmath0co absorption studies of martn - pintado et al .
h@xmath0co masers c , d , e , g , and h fall within the field imaged by martn - pintado et al .
as these authors note , all of the h@xmath0co masers occur at positions on hot nh@xmath6 shells .
the higher temperature ( t@xmath41 100 k ) region of apparent interaction between shells a and b contains the closely spaced d , g , and h masers .
martn - pintado et al .
( 1999 ) propose that the pumping mechanism for the h@xmath0co masers as well as that for the nh@xmath6 and class ii ch@xmath6oh masers must depend on the physical conditions in the hot shells . in the search for additional constraints on the h@xmath0co emission environment
, we examine the possible association of h@xmath0co masers with other molecular emission for which physical conditions are better understood . in table
[ tbl - others ] we present a summary of possible associations with h@xmath0o masers observed by mcgrath , goss , and de pree ( 2004 ) , ch@xmath6oh masers observed by houghton and whiteoak ( 1995 ) , and oh masers observed by argon , reid , and menten ( 2000 ) .
the excitation conditions for the different species are mutually exclusive , but if the masers exist in a shocked region as proposed by martn - pintado et al .
( 1999 ) for sgr b2 ( and for ngc 7538 by h03 ) , a rapid change in densities , temperatures , and velocity fields over a small linear distance is to be expected . of the nh@xmath6 masers in sgr b2 with interferometically determined positions , only masers m1 and m6
lie similarly near an h@xmath0co maser ( e ) , and both differ in velocity by @xmath4210 km s@xmath11 from the h@xmath0co maser .
the existence of h@xmath0o and ch@xmath6oh masers near the h@xmath0co f maser is significant because previously the f maser had no known maser or continuum associations ( mgp94 ) .
the h@xmath0co maser in ngc 7538 ( h03 ) and many of those reported in section 3.2 varied in intensity with timescales from years to decades .
h03 noted a common variability of some of the h@xmath0co and h@xmath0o masers in ngc 7538 .
this was further confirmed by long - term observations by lekht et al.(2003 , 2004b ) .
similarly in sgr b2 , the long - term monitoring of the h@xmath0o masers presented by lekht et al .
( 2004a , 2004c ) may indicate common variability of h@xmath0o masers with the h@xmath0co a maser .
however , variablity of an h@xmath0co maser in iras 18566 + 0408 was recently seen to occur on much more rapid timescale ( araya et al .
we present vlba+y27 images of the sgr b2 a and d h@xmath0co masers . the measured sizes ( @xmath4 au ) and brightness temperatures ( @xmath5 k ) are comparable to those found in other vlbi studies of h@xmath0co masers . however , about half of the flux density from these regions is resolved out with the vlba data .
a comparison between vla and vlba observations shows that the missing flux density exhibits a broader linewidth than the emission from the compact vlbi source .
we demonstrate quantitatively the applicability of a core - halo model for these masers .
h@xmath0o and ch@xmath6oh masers discovered near the h@xmath0co masers may indicate associations among the species , suggestive of a related origin .
the association of h@xmath0co masers in sgr b2 with hot nh@xmath6 shells proposed by martn - pintado et al .
( 1999 ) , together with the arguments for shock excitation of the maser region in ngc 7538 in h03 , provide an encouraging stepping stone toward a solution of the problem of the excitation of these poorly understood masers .
we are indebted to an anonymous referee for a very careful reading of this paper and for directing our attention to two very important references .
araya , e. , hofner , p. , linz , h. , sewilo , m. , watson , c. , churchwell , e. , olmi , l. , & kurtz , s. 2004 , , 154 , 579 araya , e. , hofner , p. , kurtz , s. , linz , h. , olmi , l. , sewilo , m. , watson , c. , churchwell , e. 2005 , , 618 , 339 araya , e. , hofner , p. , goss , w. m. , kurtz , s. , linz , h. , & olmi , l. 2006a , , 643 , l33 araya , e. , hofner , p. , sewilo , m. , linz , h. , kurtz , s. , olmi , l. , watson , c. , & churchwell , e. 2006b , ( in press ) argon , a. l. , reid , m. j. , & menten , k. m. 2000 , , 129 , 159 boland , w. & de jong , t. 1981 , , 98 , 149 bower , g. c. , backer , d. c. , sramek , r. a. 2001 , , 558 , 127 bridle , a. h. & schwab , f. r. 1999 , in synthesis imaging in radio astronomy ii . edited by g. b. taylor , c. l. carilli , and r. a. perley .
asp conference series , 180 , 371 downes , d. & wilson , t. l. 1974 , , 191 , l77 forster , j. r. , goss , w. m. , wilson , t. l. , downes , d. , & dickel , h. r. 1980 , , 84 , l1 forster , j. r. , goss , w. m. , gardner , f. f. , & stewart , r. t. 1985 , , 216 , 35p gardner , f. f. , whiteoak , j. b. , forster , j. r. , & pankonin , v. , 1986 , , 218 , 385 gaume , r. a. & claussen , m. j. 1990 , , 351 , 538 goicoechea , j. r. , rodrguez - fernndez , n. j. , cernicharo , j. 2004 , , 600 , 214 gwinn , c. r. , moran , j. m. , reid , m. j. , schneps , m. h. 1988 , , 330 , 817 hoffman , i. m. , goss , w. m. , palmer , p. , & richards , a. m. s. 2003 , , 598 , 1061 ( h03 ) hoffman , i. m. , goss , w. m. , brogan , c. l. , claussen , m. j. 2005 , , 620 , 257 houghton , s. & whiteoak , j. b. 1995 , , 273 , 1033 kobayashi , h. , ishiguro , m. , chikada , y. , ukita , n. , morita , k .-
i . , okumura , s. k. , kasuga , t. , & kawabe , r. 1989 , , 41 , 141 lazio , t. j. w. & cordes , j. m. 1998 , , 505 , 715 lekht , e. e. , munitsyn , v. a. , & tolmachev , a. m. 2003 , astronomy reports , 47 , 838 lekht , e. e. , ramrez , o. , tolmachev , a. m. , & berulis , i. i. 2004 , astronomy reports , 48 , 171 lekht , e. e. , munitsyn , v. a. , & tolmachev , a. m. 2004 , astronomy reports , 48 , 200 lekht , e. e. , ramrez , o. , & tolmachev , a. m. 2004 , astronomy reports , 48 , 965 martn - pintado , j. , de vicente , p. , wilson , t. l. , & johnston , k. j. 1990 , , 236 , 193 martn - pintado , j. , gaume , r. a. , rodrguez - fernndez , n. , de vicente , p. , & wilson , t. l. 1999 , , 519 , 667 mcgrath , e. j. , goss , w. m. , & de pree , c. g. 2004 , , 155 , 577 mehringer , d. m. , goss , w. m. , & palmer , p. 1994
, , 434 , 237 ( mgp94 ) mehringer , d. m. , palmer , p. , & goss , w. m. 1995 , , 97 , 497 palmer , p. , goss , w. m. , & devine , k. e. 2003 , , 599 , 324 reid , m. j. , schnepps , m. h. , moran , j. m. , gwinn , c. r. , genzel , r. , downes , d. , & rnnng , b. 1988 , , 330 , 809 reid , m. j. , readhead , a. c. s. , vermeulen , r. c. , & treuhaft , r. n. 1999 , , 524 , 816 rickett , b. j. 1990 , , 28 , 561 urry , c. m. , padovani , p. 1995
, , 107 , 803 whiteoak , j. b. & gardner , f. f. 1983 , , 205 , 27p l c c observing date(s ) & 2003 may 17 & 22 & 2005 october 11 , 13 , & 14 + position ( j2000.0 ) & 17 47 20.0463 , -28 23 46.587 & 17 47 19.96 , -28 22 59.8 + & 17 47 19.8562 , -28 22 12.900 & + synthesized beam & ( a ) @xmath43 mas , p.a.= @xmath44 & @xmath45 , p.a.= @xmath46 + & ( d ) @xmath47 mas ,
p.a.= @xmath44 & + flux density calibrator & & 3c286 + phase calibrator & j1745 - 283 & j1751 - 253 + bandpass calibrator & j1733 - 130 & j1733 - 130 + rest frequency & 4829.6569 mhz & 4829.6590 mhz + number channels & 128 & 63 + channel spacing & 1.953 khz & 3.052 khz + velocity resolution & 0.24 km s@xmath11 & 0.19 km s@xmath11 + center velocity & 50.6 km s@xmath11 & 75.6 km s@xmath11 & 51.4 km s@xmath11 & 74.6 km s@xmath11 + total velocity range & @xmath487.5 km s@xmath11 & @xmath485.5 km s@xmath11 + typical noise per channel & 35 mjy beam@xmath11 & 5 mjy beam@xmath11 + c r@ r@ [email protected] r@ r@ [email protected] c c c c c c a & 17&47&19&856(1 ) & @xmath4928&22&12&99(2 ) & 75.33(1 ) & 0.59(2 ) & 645(15 ) & 13(3 ) & 5(2 ) & 110(20 ) + d & & & 20&047(2 ) & & 23&46&59(2 ) & 50.07(2 ) & 0.36(5 ) & 160(20 ) & 10(4 ) & 8(4 ) & 5(40 ) + c r@ r@ [email protected] r@ r@ [email protected] c c c c a & 17&47&19&94 & @xmath4928&22&13&0 & 75.31(7 ) & 0.71(2 ) & 1900 & u + b & & & 20&04 & & 22&40&6 & 51.0(2 ) & 0.4(2 ) & 80 & u + d & & & 20&01 & & 23&47&2 & 50.1(1 ) & 0.75(5 ) & 510 & + & & & & & & & 53.9(1 ) & 0.88(9 ) & 320 & + e & & & 18&64 & & 24&24&5 & 49.1(1 ) & 0.94(4 ) & 230 & f + & & & & & & & 51.5(1 ) & 0.60(6 ) & 120 & f + f & & & 19&61 & & 22&13&5 & 76.3(2 ) & 0.3(1 ) & 40 & u + g & & & 19&57 & & 23&49&9 & 48.7(3 ) & 0.8(2 ) & 120 & u + h & & & 20&43 & & 23&46&7 & 70.6(3 ) & 0.3(2 ) & 50 & u + & & & & & & & 73.1(1 ) & 0.9(5 ) & 50 & u + & & & & & & & 74.6(1 ) & 0.8(3 ) & 70 & u , n + i & & & 24&72 & & 21&43&0 & 70.9(1 ) & 0.92(6 ) & 60 & + c c c c c a & 720@xmath4815 & 850@xmath4815 & 640@xmath4850 & 1900@xmath48400 + b & 80@xmath4815 & 150@xmath4810 & & 20@xmath4810 + d & 700@xmath4825 & 1400@xmath4830 & 55@xmath4820 & 1200@xmath48300 + e & 230@xmath4880 & 350@xmath4820 & & 330@xmath4830 + f & @xmath50 & 70@xmath4815 & & 40@xmath51 + g & @xmath52 & 320@xmath4815 & & 75@xmath53 + h & 140@xmath4860 & 350@xmath4815 & & 35@xmath4820 + i & @xmath50 & 100@xmath4815 & & 75@xmath4820 + c c c c a & 0.6 & 5000 & 0.1 + & 0.1 & 850 & 18.8 + f & 0.2 & 1700 & 9.9 + h & 0.7 & 5500 & 6.6 + & 0.7 & 5500 & 11.9 + b & 0.7 & 6000 & + d & 0.8 & 7400 & + e & 0.5 & 4200 & + f & 4.5 & 38000 & + i & 0.5 & 4200 & + a & 0.6 & 5000 & 2.5 + b & 0.6 & 5000 & 1.2 + c & 0.15 & 130 & 0.2 + & 0.05 & 40 & 0.2 + & 0.06 & 50 & 0.2 + & 0.18 & 150 & 0.2 + d & 0.07 & 60 & 0.9 + & 0.10 & 850 & 1.1 + h & 0.21 & 180 & 0.1 + & 0.24 & 200 & 2.0 + | observations of two of the formaldehyde ( h@xmath0co ) masers ( a and d ) in sgr b2 using the vlba+y27 ( resolution @xmath1 001 ) and the vla ( resolution @xmath2 ) are presented .
the vlba observations show compact sources ( @xmath3 milliarcseconds , @xmath4 au ) with brightness temperatures @xmath5 k. the maser sources are partially resolved in the vlba observations .
the flux densities in the vlba observations are about 1/2 those of the vla ; and , the linewidths are about 2/3 of the vla values . the applicability of a core - halo model for the emission distribution is demonstrated .
comparison with earlier h@xmath0co absorption observations and with ammonia ( nh@xmath6 )
observations suggests that h@xmath0co masers form in shocked gas .
comparison of the integrated flux densities in current vla observations with those in previous observations indicates that ( 1 ) most of the masers have varied in the past 20 years , and ( 2 ) intensity variations are typically less than a factor of two compared to the 20-year mean .
no significant linear or circular polarization is detected with either instrument . |
the concept of photonic crystals ( phcs ) , i.e. , artificial media with a periodic modulation of local optical characteristics , provides ways to tailor the dispersion , diffraction , and routing of electromagnetic waves @xcite . as for natural crystals ,
fundamental characteristics of the phcs are the band - diagrams , which reveal gaps where bloch waves can not propagate
. in phcs made of nonlinear materials , self - trapped localized modes in the form of the gap solitons ( gss ) may exist , as a result of the interplay of the kerr - type nonlinearity and periodic structures npc - book , krug , ostrovskaya03 , konotop .
unlike spatial bright solitons supported by the balance between the self - focusing nonlinearity and diffraction in uniform bulk media @xcite , the dispersion relation induced by the phc makes it possible to create gss in both focusing and defocusing media . combining assets of phcs and regular solitons , gss have a potential for applications to soliton - driven photonics .
new technologies enabling reconfigurable optical lattices , such as photorefractive crystals @xcite and nematic liquid crystals @xcite , also open new ways to control the dynamics of solitary waves by adjusting the lattice depth and period . while the modulational @xcite and oscillatory instabilities @xcite impose limits on the use of the gss , the stability and mobility of the gss may be enhanced in nonlinear media featuring a nonlocal response @xcite .
the nonlocal nonlinearity is important when the correlation radius of the material s response function becomes comparable to the transverse width of the wave packet @xcite .
the nonlocal nonlinearity gives rise to specific features in the soliton dynamics , including the modification of the modulational @xcite , azimuthal @xcite , and transverse @xcite instabilities .
suppression of the collapse of multidimensional solitons @xcite , a change of the character of interactions between them @xcite , formation of soliton bound states @xcite , merger of colliding solitons into a standing wave @xcite , and families of dark - bright soliton pairs @xcite were predicted recently too .
experimental observations of the nonlocal response have also been demonstrated in sundry media , including photorefractive media @xcite , nematic liquid crystals @xcite , and thermo - optical materials rotschild05,efremidis2008 , with a large tunable range of the nonlocality degree . in this work , we aim to study gs modes in the first bandgap of the model including local self - focusing and nonlocal self - defocusing nonlinearities . when the nonlocality radius is zero , we assume equal magnitudes of the self - focusing and self - defocusing terms , i.e. , complete cancelation of the nonlinearity , hence no solutions in the bandgap . for
a small degree of the nonlocality , the existence , bifurcation , and shape transitions of the emerging bright gss are analyzed . with the competing local and nonlocal interactions of opposite signs ,
the family of on - site gs modes remain stable , obeying the `` anti - vakhitov - kolokolov '' criterion antivk , for the case that the nonlocal perturbation of the refractive index is small , but become unstable near the upper edge of the bandgap .
however , the bifurcation generating the gss near the other edge of the bandgap features an inverted slope of the bifurcation curve for a relatively small degree of the nonlocality , in which case the on - site gss are unstable .
the off - site gs family features single - humped , double - humped , and flat - top profiles for different degrees of the nonlocality .
we also investigate the situation in the space of the soliton s propagation constant and power , for the varying nonlocality degree , in order to illustrate the mechanism underlying the complex bifurcation pattern and related cutoff effects ( termination of some solution branches ) .
using results reported in this work , we discuss the possibility to design gs - based signal - processing schemes by dint of manipulating the nonlocal interactions .
in addition to optics , gss of matter waves have also been theoretically studied @xcite and experimentally created @xcite in bose - einstein condensates ( becs ) formed by atoms with repulsive interactions , trapped in optical - lattice ( ol ) potentials .
in addition to the known contact interaction in the becs of alkali - metal atoms , the interaction of chromium atoms , @xmath0cr , includes a dipole - dipole interaction , which is intrinsically anisotropic and nonlocal .
the condensate of @xmath0cr was created and investigated using magnetic @xcite and all - optical @xcite traps , see also review @xcite . by adjusting the orientation of the dipoles , one can effectively control the nonlocal dipole - dipole interactions .
for the dipolar bec trapped in ols , the competition between the contact and long - ranged dipole - dipole interactions not only dramatically change the band structures of nonlinear bloch waves @xcite , but also modifies families of matter - wave solitons @xcite .
the rest of the paper is organized as follows . in sec .
[ sec2 ] , the model including the competing local and nonlocal nonlinearities is described .
properties of the on - site and off - site gs families supported by the local self - focusing nonlinearity are recapitulated to show a transition of mode profiles in the first bandgap .
results produced by the interplay of the local self - focusing and nonlocal self - defocusing nonlinearities for on - site and off - site gs families are reported in sections [ sec4 ] and [ sec5 ] , respectively . tracing the change of the corresponding gs shapes in the parameter planes , we explain the character of the corresponding bifurcation and identify a possible control mechanism for the optical - signal processing .
section [ sec6 ] concludes this work .
considering a wave packet propagating along the @xmath1 axis in the nonlinear phc structure , we assume that the embedded medium gives rise to two kinds of the nonlinearity simultaneously .
this system is modeled by the modified nonlinear schrdinger equation @xcite , @xmath2where @xmath3 is the slowly varying amplitude of the electric field , @xmath4 is the periodic potential , and a perturbation of the refractive index , @xmath5 , accounts for the diffusive nonlinear response with the nonlocality degree ( which scales as a squared nonlocality radius ) designated by parameter @xmath6 .
sign parameters @xmath7 and @xmath8 correspond to the self - defocusing and self - focusing nonlinearities , respectively .
below , we fix @xmath9 and @xmath10 for a system with the nonlocal self - defocusing and local self - focusing nonlinearities . in the case of the local self - focusing kerr nonlinearity , depicted by black and gray lines , respectively .
the propagation constant of each mode is indicated by labels a through f in panel ( g ) , where the relations between the propagation constant and power are plotted by the solid and dashed curves , respectively , for the on - site and off - site modes .
panel ( i ) shows the blow - up of the area enclosed by the dashed box in ( g ) . here and in other figures ,
shaded areas represent bloch bands bordering the first finite bandgap.,width=313 ] stationary solutions with propagation constant @xmath11 are looked for as @xmath12 . for periodic potential @xmath13 with @xmath14 fixed by rescaling , the linearized version of eq .
( [ eq : gp1 ] ) gives rise to the bandgap structure , with the first finite bandgap being @xmath15 . in this bandgap
, gss can exist in form of on - site and off - site modes being supported solely by the local nonlinearity , i.e. , with @xmath16 bifur - gap-2004 . for the self - focusing kerr nonlinearity , we display examples of both the on - site and off - site gs solutions in fig .
fig : mode - d0 .
the relationships between @xmath17 and the power of these modes , @xmath18 , are shown in the form of the bifurcation curves in fig .
[ fig : mode - d0](g ) , with labels a through f referring to typical gs mode profiles in subplots ( a - f ) . under the self - focusing nonlinearity , these gss bifurcate from the upper edge of the first bandgap ( in other words , from the lower edge of the second finite bloch band ) , where the spatial - dispersion law features the anomalous sign @xcite .
in the entire first bandgap , the odd on - site gs modes , shown in fig . [
fig : mode - d0](a - c ) , feature two major peaks within one lattice cell .
tails of these modes become conspicuous when the propagation constant moves close to the upper edge of the bandgap . on the other hand ,
a shape transition is demonstrated by the off - site even gs modes . near the lower edge of the bandgap
, the gs solution has a single major peak coinciding with a local maximum of the periodic potential , as shown in fig .
[ fig : mode - d0](d ) . by tracing the variation of the propagation constant and power along the relation for these off - site modes shown by the dashed line of fig .
[ fig : mode - d0](g ) to point f , it is seen in fig .
[ fig : mode - d0](f ) that the center of the modal profile breaks into two peaks .
the double - peaked solution is formed due to the balance between the repulsive potential barrier and self - trapping induced by the kerr nonlinearity , as the characteristic self - trapping length becomes larger than a half of the lattice period ( roughly equivalent to the width of the barrier potential ) . in between
, we observe a smooth shape transition of the gs from the single - humped shape in fig .
[ fig : mode - d0](d ) to a nearly flat - top one in fig .
[ fig : mode - d0](e ) , and , finally to the double - peaked mode in fig .
[ fig : mode - d0](f ) .
the off - site gss exist when the nonlinear self - trapping is stronger than the repulsion induced by the potential barrier , giving rise to an effective a potential well holding the localized modes , as seen in figs .
fig : mode - d0(d - f ) . on the other hand , the effect of the lattice potential is stronger than that of the nonlinearity in the case of the on - site modes , which , together with the contribution of the gradient energy , determines their shapes in fig .
[ fig : mode - d0](a - c ) .
when the refractive index correction gets weaker in accordance with the power reduction , the concave lattice potential create a potential barrier to tailor and split the wave function into a form of the states that similar to a binding profile from two on - site modes in fig .
[ fig : mode - d0](b ) .
in this section , we introduce the self - defocusing nonlocal nonlinearity , setting @xmath19 in eq .
( [ eq : gp1 ] ) . obviously , in this case the total nonlinearity cancels out to zero in the limit of @xmath20 . at @xmath21 ,
the overall nonlinearity is a self - focusing , because the diffusive nonlocal kernel produces a spatially wider and less intensive perturbation of the nonlinear refractive index , in comparison with that corresponding to the local nonlinearity .
then , similar to the situation in the linear model outlined above , one may expect the corresponding gs modes to bifurcate from the upper edge of the first bandgap , where the effective spatial dispersion is anomalous . in fig .
[ fig : cn - onsite ] , we show examples of on - site gss supported by the competing local focusing and nonlocal defocusing nonlinearities for different degrees of non - locality , @xmath22 , and the case of @xmath23 ( the local nonlinearity only)for the comparison .
( gray lines ) , for different degrees of the non - locality @xmath24 , and the case of @xmath25 ( local only ) .
the corresponding values of propagation constant @xmath26 are labeled ( a - e ) in panel ( f ) .
the dashed line in ( f ) is the asymptotic curve for @xmath27 . , width=313 ] for different degrees of the nonlocality , the profiles of the on - site gs solutions vary slightly , remaining similar to their counterparts in the local model .
on the other hand , fig .
[ fig : cn - onsite](f ) demonstrates that the power required for the formation of the gss is higher in the model with the competing nonlinearities than in the local one because the net nonlinearity is effectively reduced by the competition of the nonlocal nonlinear response with the local kerr term . in the limit of the ultimate nonlocality , @xmath28
, the @xmath29 curve converges to that in the local model ( with @xmath23 ) .
the latter feature is explained by the fact that , in this limit ( which corresponds to the model of the so - called `` accessible solitons '' @xcite ) , the nonlocal nonlinear response amounts to a weak constant background , which shifts the propagation constant by the amount proportional to @xmath30 pra-10-yylin . .
the inset shows the blow - up of stability analysis along with the corresponding power - propagation constant relations.,width=302 ] in fig .
[ fig : cn - onsite](f ) , the bifurcation curve shows an abrupt change of the slope both for @xmath31 and @xmath32 ( in the latter case , near point e ) . in a small region to the left of this point , the slope changes its sign to @xmath33 , and the corresponding on - site gss are unstable .
it is known that the bifurcation curves for gs families supported by the local self - focusing nonlinearity obeys an inverted `` anti - vakhitov - kolokolov '' criterion @xcite , @xmath34 , under which the on - site gss are stable ( see also ref .
@xcite ) . to elucidate
, we analyzed the stability of the numerically found gs families by considering small - perturbation modes and calculating their eigenvalues .
in specific , we show in fig .
[ fig : cn - onsite - stab ] the linear stability analysis spectrum over the entire first band gap regime .
it is clearly illustrated in the inset of fig .
[ fig : cn - onsite - stab ] , unstable small - perturbation eigenmodes are obtained only within the region where the power dependence @xmath33 and therefore confirms the inverted `` anti - vakhitov - kolokolov '' criterion .
the evolution of these on - site gap solitons generated by direct numerical simulations further shows the collapse of on - site solitons falling out of the `` anti - vakhitov - kolokolov '' regime . when the propagation constant of the gs is close to the upper edge of the bandgap ( on the right - hand side of the slope - change point ) , the slope of the @xmath35 becomes negative again . in that case , the gs is broad , spanning a few lattice periods , and resembles _ gap wave _ modes @xcite , as seen in fig .
[ fig : cn - onsite](e ) .
the abrupt slope change smooths out with the increase of the nonlocality degree , disappearing at @xmath36 , due to the fact that the nonlocal perturbation of the refractive index becomes small for the strong nonlocality . for a sufficiently small degree of the nonlocality , @xmath6 , fig .
fig : cn - onsite shows that the power required to form the gss increases with the decrease of @xmath6 , diverging at @xmath37 . in the regime of the weak nonlocality
, the effective competing nonlinearity can be approximated , to the first order in @xmath6 , as @xcite @xmath38 .
an asymptotic curve based on this approximation is shown by the dashed line in fig . [
fig : cn - onsite](f ) , to illustrate the bifurcation of the on - site gss near the upper edge of the first bandgap .
next , we aim to study off - site gss in the first finite bandgap under the action of the competing nonlinearities .
as mentioned in section [ sec2 ] , the profile of the off - site gs solutions changes from single - humped to double - humped as propagation constant @xmath39 approaches the upper edge of the bandgap . to indicate the change of the profile caused by the introduction of the competing nonlocal nonlinearity , in fig .
fig : cn - offsite(c ) we use thin and thick lines to distinguish portions of the @xmath40 curves representing such single- and double - humped profiles . again , in the limit of the strong nonlocality , the correction to the refractive index induced by the nonlocal nonlinearity is widely spread in the space and very small with respect to the effect of the kerr nonlinearity , which makes the competition negligible .
for example , for @xmath41 [ see fig .
[ fig : cn - offsite](c ) ] , both the @xmath40 curve and the corresponding shape transition of the off - site gss are close to their counterparts in the local model .
a smooth transition of the gs solutions from the single - peaked shape to a flat - top one , and then to the double - peaked ( in - phase ) shape can be traced , in the latter case .
however , for a smaller degree of the nonlocality , which makes the self - defocusing nonlocal response comparable to the kerr nonlinearity , the analysis reveals the existence of more than one branches of the off - site gss . for small values of @xmath6 , such as @xmath42 shown in fig .
fig : cn - offsite(c ) , one branch ( the bold red curve in the figure ) extends continuously from the upper to lower edge of the bandgap .
the modal profile for this branch remains double - humped , such as the one shown by dashed lines of fig .
[ fig : cn - offsite](a ) and ( b ) for markers b and d in ( c ) , respectively . besides this branch of the double - humped gs modes
, there is a separate branch representing solutions with a lower formation power and single - humped profile , as shown by solid lines in fig .
[ fig : cn - offsite](a ) and ( b ) for markers a and c in panel ( c ) . and @xmath43 , are shown in ( c ) to indicate the cases with and without the new branch.thin and thick portions of @xmath44 curves pertain to the solitons with single- and double - peak profiles , respectively.,width=313 ] thus , the single- and double - peaked gss , which constitute parts of the single gs family in the local model , split into two disjoint families in the model with the appreciable competition between the local and nonlocal nonlinearities . in the intermediate case , corresponding to a moderate degree of the nonlocality , such as for @xmath45 shown in fig .
[ fig : cn - offsite-2](b ) , the @xmath40 curve for the double - humped gss break into two segments : one starts from the lower edge of the bandgap and ends at a cutoff point corresponding to marker b in the figure , and the other starts from the upper edge and ends at another cutoff point , which is designated by marker @xmath46 .
the cutoff points may be accounted for by bifurcations involving additional higher - order modes , which we did not aim to find in this work dealing with fundamental on- and off - site gss .
another @xmath40 branch chiefly represents the single - humped gss , but it also contains a portion to the right of point a , which corresponds to double - humped modes . the inset of fig .
[ fig : cn - offsite-2](b ) clearly shows that there are three branches of double - humped modes .
the first one bifurcates from the upper edge of the bandgap and ends at the point marked @xmath47 ; the second one extends from the lower edge of the bandgap and terminates at point labeled by @xmath48 ; the final branch bifurcates as the flat - top solution from the point marked the asterisk ( @xmath49 ) and terminates at the site marked @xmath50 .
the first two branches that bifurcate from either edge of the bandgap abruptly terminate inside the bandgap , where the characteristic width of the nonlinear response is larger than or comparable to half the lattice period , which makes the balance between the nonlinear and lattice - induced effects impossible .
the last branch , which starts as the flat - top mode , ends due to the divergence of the total power as a result of vanishing nonlinearity , similar to cutoff considered in ref .
even though off - site gss are in general unstable both in local @xcite and nonlocal @xcite nonlinearities despite the inverted `` anti - vakhitov - kolokolov '' criterion , the corresponding instability growth rate is proportional to the gs s power after a certain threshold value @xcite . due to the unstable nature of off - site gss
, we study the instability of gss by the linear stability analysis for the branches off - site gss and identify the final state of these off - site gss by beam propagation simulation .
the linear stability spectrum in fig . [ fig : cn - offsite - stab ] shows the amplitude growth rate of the small - perturbation eigen - modes found upon those off - site gs s revealed previously . for a smaller degree of non - locality , @xmath42 , illustrated in fig .
[ fig : cn - offsite - stab](a ) the two distinct branches are unstable and the corresponding eigen - modes upon single - peaked gss have higher amplitude growth rate which diverges near its cutoff point in the band gap region .
the gss on these two branches collapse fast as they propagates due to a larger growth rate . likewise , in fig . [
fig : cn - offsite - stab](b ) when a moderate degree of non - locality , @xmath45 , is considered , the single - peaked gss branch acquires a stronger instability than that of the two separated double - peaked gss branches . the two double - peaked gss branches ,
though both feature the worst instability near their cutoff points in the band gap region , reflect very different relations to the corresponding gs s power , which we believe is stemmed from the cutoff of the branches .
the beam propagation simulations further illustrate that gss of the single - peaked branch and double - peaked modes with a higher energy fall into collapsed states .
nevertheless , mode conversions from unstable off - site gss into stable on - site gss are observed for the branch plotted in blue in fig .
[ fig : cn - offsite-2](b ) .
three examples of beam propagation simulations are illustrated , resulting in either collapsed states fig .
[ fig : cn - offsite - porop ] ( a ) and ( b ) or a mode conversion behavior fig .
[ fig : cn - offsite - porop ] ( c ) .
information such as mode transition or conversion is beyond what linear stability analysis may reveal .
even though it is also believed that interesting dynamical behavior associated with the gss can be delineated through a direct beam propagation simulation , yet to be more focused , a thorough investigation of propagation behavior goes beyond the scope of this work .
curves for the nonlocality degree @xmath45 are shown .
the thin and thick lines distinguish portions corresponding to single- and double - peaked modes , respectively .
inset in ( b ) is a blow - up of the region around the marked points.,width=313 ] and ( b ) @xmath45 .
the color and line style to distinguish each branch is as is defined in the corresponding @xmath44 curves in in fig .
[ fig : cn - offsite](a ) and fig .
[ fig : cn - offsite-2](b ) , respectively .
, width=313 ] three examples of the beam propagation simulations for off - site gs solutions shown in fig .
[ fig : cn - offsite-2](a ) , in which modes a , b and c correspond to intensity plots in ( a ) , ( b ) , and ( c ) , respectively.,width=351 ] the nonlocal nonlinearity competing with the local kerr term not only reduces the strength of the nonlinearity but also tailors the effective nonlinear response to induce binding forces outside of the lattice - potential barriers , which is a more favorable environment for the existence of double - peaked modes .
therefore , the three gs branches outlined above are formed owing to the interplay of the potential barrier created by the lattice potential and the binding potential induced by the effectively _ reduced _ kerr nonlinearity . in this case , the gs solutions belonging to the branch originating from the lower edge of the bandgap have a larger amplitude and are tighter localized . the width of the corresponding response range of the self - defocusing nonlocal nonlinearity is larger than that of the kerr response .
such a double - peaked branch can not exist in the strong - nonlocality limit , and we numerically find that values of the nonlinearity degree supporting this branch are bounded by @xmath51 .
as the power decreases ( the propagation constant increases ) , the widening of the the gs mode makes the overall nonlinear response effectively local , suppressing the capability of the nonlocal nonlinearity to tailor its response to a shape necessary for supporting the solitons .
then , when the gs width becomes smaller than or comparable to half the lattice period , the double - peaked modes cease to exist because the balance between the lattice potential and nonlinearity - induced perturbation of the refractive index supports only single - peaked modes . to present a clear description of shape transitions for the off - site gss , we replot the relationship of power @xmath52 versus the nonlocality degree , @xmath6 , for a fixed propagation constant @xmath17 in fig .
[ fig : p - d ] . for a smaller propagation constant , such as @xmath53 in fig .
[ fig : p - d](a ) , the power necessary to support a single - humped off - site gs is always lower than that of its double - humped counterpart , for all values of @xmath6 .
moreover , above a critical degree of the nonlocality , @xmath54 in this case , only a single - humped gs can be found . moving @xmath17 into the center of the bandgap for instance , taking @xmath55 in fig .
[ fig : p - d](b ) the critical degree of the nonlocality reduces to @xmath56 , and , above another critical value , @xmath57 , the single - humped mode transforms into a double - humped one , as in the local model . increasing the value of the propagation constant to @xmath58
, the two @xmath59 curves merge at @xmath60 in in fig .
fig : p - d(c ) . for a larger value of the propagation constant , such as @xmath61 in fig .
[ fig : p - d](d ) , the two curves intersect at a critical value @xmath62 . above this critical point , the power for the double - humped off - site gs becomes lower than for a single - humped one . in this case
, the profile of the off - site gs mode can be switched from single - humped into double - humped by adjusting the nonlocality degree , @xmath6 .
, versus the degree of the nonlocality , @xmath6 , at fixed values of the propagation constant : @xmath63 ( a ) , @xmath64 ( b ) , @xmath65 ( c ) , and @xmath66 ( d ) , respectively . the inset in ( c ) is the blow - up of the region in the vicinity of the merger point .
, width=313 ] to facilitate the understanding of the present picture , we consider the plane of the propagation constant @xmath17 and power @xmath52 for the gs solutions in fig .
[ fig : phase - diagram ] , varying the nonlocality degree @xmath6 .
we start by tracing the evolution of the point of the transition from single - peaked to the double - peaked shape , marker by the asterisk ( @xmath67 ) in fig .
[ fig : cn - offsite-2](c ) . in the absence of the competing nonlocal nonlinearity ,
i.e. , at @xmath23 in eq .
( [ eq : gp1 ] ) , the transition point is @xmath68 , labeled by a in fig . [
fig : phase - diagram ] , which also corresponds to the limit of @xmath28 .
as the nonlocality degree drops to a critical value , @xmath31 at point c @xmath69 , the transition point ceases to exist ( i.e. , only sharply peaked single - humped modes are supported by the system ) , merging into to the end point of the doubled - humped - mode branch , marked by @xmath50 in fig .
[ fig : cn - offsite-2 ] . increasing the nonlocality degree from @xmath31 at point c , the end point @xmath50 in fig .
[ fig : cn - offsite-2 ] merges into the other end point @xmath70 in fig .
[ fig : cn - offsite-2 ] at @xmath71 .
the latter merger happens at point b in fig .
fig : phase - diagram .
this is the end point of the family of the double - peaked modes [ an example corresponds to the point marked by @xmath47 in the inset of fig . [
fig : cn - offsite-2](b ) ] which originates from point b in fig .
[ fig : phase - diagram ] at the critical nonlocality @xmath72 , and extends toward point d , which corresponds to @xmath73 , where it merges into a new branch of the double - peaked modes emerging ( as long as @xmath51 ) from the lower edge of the bandgap at point @xmath74 . ) and power ( @xmath52 ) , for the gap solitons , as different values of the nonlocality degree , @xmath6 .
the inset is a blowup of the gray region .
route ac corresponds to the transition point marked by @xmath67 in fig .
[ fig : cn - offsite ] ; routes bc and bde correspond to the double - peaked modes marked by @xmath70 and @xmath75 in fig .
[ fig : cn - offsite-2 ] , respectively ; the route above point c corresponds to the single - peaked solution .
the marked points are a with @xmath76 , and b , c , d and e with @xmath72 , @xmath77 , @xmath78 , and @xmath79 , severally.,width=302 ]
in this work , we aimed to study gs ( gap - soliton ) solutions in the first finite bandgap of the periodic potential , with the nonlinearity represented by the competing local self - focusing and nonlocal self - defocusing terms .
the two terms are balanced so that , in the limit of the zero nonlocality radius , they exactly cancel each other .
while keeping the effective interaction self - attractive , the existence , stability , and bifurcation for on - site and off - site modes were analyzed numerically .
due to the opposite signs of the local and nonlocal nonlinearities , an increased power was required for the formation of both the on - site and off - site gss .
the competing nonlinearities induce a region where stable on - site modes obeying the `` anti - vakhitov - kolokolov criterion '' near the upper edge of the bandgap region become unstable . for unstable off - site gs modes , which remain unstable under competing local and nonlocal nonlinearities , a complex bifurcation pattern with cutoff points
was found and explained in terms of the transitions between the single - humped , flat - top , and double - humped shapes . by tracing the evolution with the change of the nonlocality degree
, we have shown that it is possible to switch different off - site gs modes by manipulating the nonlocal interaction against the local kerr nonlinearity .
this work was partly supported by the national science council of taiwan with contrasts nsc 95 - 2112-m-007 - 058-my3 , nsc 95 - 2120-m-001 - 006 and nsc 98 - 2112-m-007 - 012 .
abdullaev , b.b .
baizakov , s.a .
darmanyan , v.v .
konotop , and m. salerno , phys .
a * 64 * , 043606 ( 2001 ) ; i. carusotto , d. embriaco , and g. c. la rocca , _
ibid_. * 65 * , 053611 ( 2002 ) ; e.a .
ostrovskaya and yu.s .
kivshar , _ phys .
lett . _ * 90 * , 160407 ( 2003 ) .
a. griesmaier , j. werner , s. hensler , j. stuhler , and t. pfau , _ phys .
lett . _ * 94 * , 160401 ( 2005 ) .
j. werner , a. griesmaier , s. hensler , j. stuhler , t. pfau , a. simoni and e. tiesinga , _ phys .
lett . _ * 94 * , 183201 ( 2005 ) .
j. stuhler , a. griesmaier , t. koch , m. fattori , t. pfau , s. giovanazzi , p. pedri , and l. santos , _ phys .
lett _ , * 95 * , 150406 ( 2005 ) .
a. griesmaier , j. stuhler , and t. pfau , _ appl .
b _ * 82 * , 211 ( 2006 ) . q. beaufils , r. chicireanu , t. zanon , b. laburthe - tolra , e. marchal , l. vernac , j .- c . keller , and o. gorceix , phys .
a * 77 * , 061601(r ) ( 2008 ) . | we analyze the existence , bifurcations , and shape transformations of one - dimensional gap solitons ( gss ) in the first finite bandgap induced by a periodic potential built into materials with local self - focusing and nonlocal self - defocusing nonlinearities .
originally stable on - site gs modes become unstable near the upper edge of the bandgap with the introduction of the nonlocal self - defocusing nonlinearity with a small nonlocality radius .
unstable off - site gss bifurcate into a new branch featuring single - humped , double - humped , and flat - top modes due to the competition between local and nonlocal nonlinearities .
the mechanism underlying the complex bifurcation pattern and cutoff effects ( termination of some bifurcation branches ) is illustrated in terms of the shape transformation under the action of the varying degree of the nonlocality .
the results of this work suggest a possibility of optical - signal processing by means of the competing nonlocal and local nonlinearities . |
osmosis is the physical phenomenon in which a concentration difference of impermeant solute molecules across a semi - permeable membrane produces a difference in solvent density , and in pressure , across the membrane .
this has enormously important consequences - osmosis forms the basis of the transport of ions across cell membranes @xcite , the regulation of blood salt levels by the kidneys @xcite , uptake of water by plants @xcite , technologies for kidney dialysis @xcite and clean power production @xcite , and many other processes .
osmosis also lies at the heart of many important phenomena in chemical physics including the donnan effect @xcite , the depletion interaction @xcite and , recently , mechanisms for generating self - propelled colloidal particles @xcite .
the cornerstone of our understanding of osmosis is the van t hoff relation , which states that a solute concentration difference , @xmath0 , leads to an equilibrium osmotic pressure difference @xmath1 @xcite .
remarkably , this relation predicts that the osmotic pressure difference is the same as the pressure of an ideal gas at concentration @xmath0 , regardless of the molecular nature of the solute and solvent molecules .
gibbs showed that the van t hoff relation can be derived by setting the chemical potential of the solvent equal across the membrane @xcite ; the thermodynamic origin of the osmotic pressure difference is then the entropy of mixing between solute and solvent .
the van t hoff relation only holds for low solute concentrations ; at higher concentrations , deviations from this `` ideal '' behaviour can be used as a diagnostic of solute - solute interactions , such as ion pairing @xcite . from a thermodynamic point of view ,
osmosis is thus well understood . however , osmotic effects are also important in non - equilibrium systems @xcite , for which a thermodynamic description is not valid . for systems that are out of equilibrium , a clear picture of the underlying molecular dynamical mechanisms
is an essential prerequisite for understanding osmotic phenomena . yet
such a picture is largely lacking , even though the dynamical basis of osmosis has been the subject of over 100 years of discussion , since the seminal work of van t hoff .
relevant factors may include bombardment of the membrane by solute molecules , diffusion of solvent across the membrane driven by its density difference , specific types of interaction between solute and solvent molecules and pulling of solvent molecules across the membrane in the wake of solute - membrane collisions @xcite .
while the overall picture remains unclear , progress has been made in understanding the dynamical basis of osmosis in specific systems . in an important contribution
, brady @xcite showed that for a system of colloidal solutes suspended in a coarse - grained solvent , the osmotic pressure can be derived from a purely mechanical , hydrodynamic langevin description of colloidal motion ; here , hydrodynamic interactions between colloids , mediated by solvent , play a central role .
for systems with explicit solvent particles , guell and brenner @xcite showed that the van t hoff relation can be derived from first principles by taking into account all the forces involved ; here , the role of membrane - solute interactions is emphasized .
importantly , while a thermodynamical description of osmosis can not depend on the details of the system s dynamical rules ( e.g. the interparticle interactions , membrane geometry , etc . ) , the nature of the underlying dynamical mechanisms may be sensitive to these details , and could be different for different osmotic systems . indeed , detailed models of osmotic water transport through membrane pores in ionic solutions have highlighted the importance of specific solvent - membrane and solvent - ion interactions in these systems @xcite , while the unusual characteristics of osmotic water flow through carbon nanotube membranes can be described by 1d stochastic hopping models @xcite .
other studies have shown that partial solute penetration into membrane pores can strongly influence solvent flow @xcite , and that , for ionic solutes , non - trivial coupling can arise between the transmembrane electric field and osmotic flow @xcite . in this paper , we seek to provide a clear `` minimal '' picture of the dynamical mechanisms underlying osmosis , in a system which is as simple as possible . murad and powles @xcite , and itano _ et al _ @xcite , have demonstrated that osmosis can be achieved in systems in which solute and solvent particles have identical lennard - jones interactions , while luo and roux @xcite have used an external potential , invisible to the solvent , to mimic a semi - permeable membrane for ionic systems . here , we combine these elements to create a `` minimal model '' osmotic system with purely repulsive particles , and with no solvent - membrane interactions , for which we carry out a detailed study of the dynamics of solvent and solute particles in the osmotic steady state . for this system , we are able to derive a simple theoretical expression for the osmotic pressure in terms of solute - solute and solute - solvent interactions . analyzing the forces on the solvent particles in our system leads us to an intuitive model in which the solvent density and pressure gradients in osmotic equilibrium are maintained by a balance between an outward solvent flux generated by the osmotic pressure gradient and an inward diffusive flux .
although more complex mechanisms may be at play in real osmotic systems , they are not required to describe the fundamental physics of osmosis in this minimal model system .
this work should provide a basis on which to build future descriptions of osmotic phenomena in both equilibrium and non - equilibrium systems .
our model system is designed to reproduce the essential physics of osmosis , while eliminating as far as possible complications due to the specific chemical nature of the solutes , solvent and the semi - permeable membrane . in our system solute and solvent particles
interact with each other indistinguishably , and there are no solvent - membrane interactions . our simulation setup is illustrated in figure [ fig : potential ] .
the simulation box is partitioned into `` solution '' and `` solvent '' compartments ; the solute particles are confined within the central solution compartment by an external potential of the form @xmath2 ( where @xmath3 is the perpendicular distance between a solute particle and the boundary of the compartment , and @xmath4 is the particle diameter ) . the solvent particles do not experience the confining potential and are free to move throughout the simulation box .
the confining potential therefore acts like a semi - permeable membrane .
both solvent and solute particles are of unit mass @xmath5 and interact via identical , repulsive , weeks , chandler , andersen ( wca ) interactions ( @xmath6 $ ] if @xmath7 and zero otherwise ) , with parameters @xmath8 and @xmath9 @xcite ( i.e. our units of energy and distance are @xmath10 and @xmath4 respectively ) .
the system is simulated using molecular dynamics , with the velocity verlet algorithm @xcite and timestep 0.001 ( in reduced units @xcite , ; in our simulations @xmath5 , @xmath4 and @xmath11 are all set to unity . ] ) , combined with a nos - hoover thermostat @xcite . in total , our system contains 5000 particles ( solute plus solvent ) in a periodic , cubic simulation box of size @xmath12 , so that the total particle density @xmath13 ( the packing fraction @xmath14 ) . in this paper , we focus on the osmotic steady state ; we therefore allow the system to equilibrate thoroughly before data is collected .
we measure the local pressure in the solution and solvent compartments using the method of planes @xcite .
as discussed in appendix [ app : mop ] , this constitutes a direct measurement of the kinetic and interaction components of the momentum flux across a local plane . the osmotic pressure is then computed by taking the difference between the pressures in the two compartments .
we could also have measured the osmotic pressure directly from the force exerted by the solute particles on the membrane ( as done by luo and roux @xcite ) , or using virial expressions for the local pressure @xcite ; these methods give identical results . to compute the concentration of solute particles , @xmath15
, we need to define the volume of the solution compartment .
this is non - trivial , because the confining potential is smooth .
we define the solution volume by matching the pressure - density relation for a gas of solute particles , confined in the solution compartment , in the absence of solvent , with that of a system of wca particles simulated in a periodic box ( see appendix [ app : mop ] ) .
as a function of solute concentration @xmath15 ( both in units of @xmath16 ) , in the solution and solvent compartments ( circles and squares respectively ) .
the dashed lines show theoretical predictions based on the carnahan - starling equation of state ( see appendix [ app : cs]).,scaledwidth=40.0% ] we first verify that osmosis indeed occurs in our model system by measuring the steady - state local density profiles in the solution and solvent compartments .
figure [ fig : density1 ] shows density profiles , taken through the middle of our simulation box . as expected , the _ total _ particle density , solute plus solvent ( figure [ fig : density1]a ) , is higher in the solution than in the solvent compartment .
in contrast , the _ solvent _ density is lower in the solution than in the solvent compartment ( figure [ fig : density1]b ) .
this density imbalance increases linearly with the concentration of solute , @xmath15 , as shown in figure [ fig : density2 ] .
these results are in good agreement with thermodynamic predictions , obtained by approximating the wca particles as hard spheres and setting the solvent chemical potential , obtained from the carnahan - starling equation of state , equal in the two compartments ( dashed lines in figure [ fig : density2 ] ; for details of the calculations see appendix [ app : cs ] ) . figure [ fig : density2 ] also demonstrates that over the parameter range considered here , the particle density in the solvent compartment remains virtually unaffected by changes in the solute concentration , confirming that the solvent compartment is large enough to be regarded as a reservoir . the osmotic pressure , @xmath17 ( i.e. the pressure difference between the solution and solvent compartments ) , is shown in figure [ fig : dp - vs - conc ] as a function of the solute concentration , @xmath15 . at low solute concentration ,
our simulation results ( circles ) are in good agreement with the van t hoff relation ( dotted line ) . at high solute concentration ( @xmath18 , corresponding to solute packing fraction greater than @xmath19 ) , the osmotic pressure exceeds that predicted by the van t hoff relation , as expected in a system with repulsive solute - solute and solute - solvent interactions .
interestingly , however , the osmotic pressure in our system is significantly lower than the pressure that would be obtained for a gas of wca particles at density @xmath15 ( dashed line in figure [ fig : dp - vs - conc ] ; computed using the carnahan - starling equation of state ) . thus a nave picture in which one treats
the solution simply as a `` solute gas '' , ignoring the solvent , does not correctly account for the osmotic pressure .
figure [ fig : dp - vs - conc ] also shows the osmotic pressure predicted by a full thermodynamic calculation , for solvent and solute , using the carnahan - starling equation of state ( dot - dashed line ) ; this prediction agrees well with our simulations , as expected for a system of particles with hard sphere - like interactions .
( in units of @xmath20 ) as a function of solute concentration @xmath15 ( in units of @xmath16 ) .
in the main plot , the symbols show simulation results .
the circles show direct measurements of @xmath17 in our simulations , computed using the method of planes ( see appendix [ app : mop ] ) .
the squares show @xmath17 computed from our simulations using eq .
( [ eq : vhpluscorr ] ) .
the statistical errors on the symbols are @xmath21 for @xmath22 ( circles and squares ) and @xmath21 for @xmath17 ( squares only )
i.e. approximately the size of the symbols .
these errors arise mainly from uncertainty in the position of the solution boundary , as discussed in appendix [ app : mop ] .
the lines show theoretical predictions .
the dotted line shows the van t hoff relation , @xmath23 . the dashed line shows the pressure of a system of hard spheres at density @xmath15 , computed using the carnahan - starling equation of state .
the dot - dashed line shows the osmotic pressure predicted by a full thermodynamic calculation , including solute and solvent , using the carnahan - starling equation of state . in the inset ,
the circles are the same as in the main plot , while the triangles show the prediction of our simple hopping model ( eq.([eq : hopping1 ] ) ) @xmath24}$ ] ( where @xmath25 and @xmath26 are the average solvent densities in the solution and solvent compartments , respectively ) .
, scaledwidth=40.0% ] standard derivations of the osmotic pressure ( including the dot - dashed line in figure [ fig : dp - vs - conc ] ) rely on thermodynamic principles
i.e. on equality of the chemical potential across the membrane .
in contrast , our aim in this paper is to investigate the _ mechanical _ origins of osmosis .
it is thus of interest to derive expressions for the osmotic pressure purely from mechanical principles , without recourse to thermodynamics . in appendix [ app : virial ] we present one such derivation .
starting from the clausius virial relation for the solute particles , we derive a ( to our knowledge new ) approximate expression for the osmotic pressure in terms of the forces of interaction between the solute and solvent particles : @xmath27 here , @xmath28 is the volume of the solution compartment ( defined as described above ) , @xmath29 is the position of particle @xmath30 and @xmath31 is the force exerted on particle @xmath30 by particle @xmath32 . the first term in the brackets sums over all pairs of solute particles ( avoiding double counting )
while the second term sums over all solute - solvent pairs ( note that here particle @xmath30 denotes the solute and @xmath32 the solvent ; the solvent may be inside or outside the solution compartment ) .
since the starting point of this derivation , the clausius virial relation , follows from newton s equations of motion in a system at steady state ( see appendix [ app : virial ] ) , eq .
( [ eq : vhpluscorr ] ) amounts to a purely mechanical description of the osmotic pressure in our system .
figure [ fig : dp - vs - conc ] ( squares ) shows that eq .
( [ eq : vhpluscorr ] ) is indeed in good agreement with the direct measurement of @xmath17 from our simulations ( circles ) , over the full range of solute concentrations tested .
expression ( [ eq : vhpluscorr ] ) makes an interesting connection with `` effective single - component '' descriptions of osmotic systems , as used , for example , for colloidal dispersions . here
, one aims to coarse - grain the system of solute and solvent particles , representing it by a single - component fluid of solute particles , which interact via an effective potential that takes into account the effects of the solvent .
this works well for colloidal dispersions where the colloids ( solutes ) are orders of magnitude larger than the solvent molecules and the effective interactions are well - represented by a pairwise intercolloidal potential @xmath33 . here , by analogy with atomic systems the osmotic pressure , @xmath34 , is written as @xmath35 , where @xmath36 is the number density of colloidal particles and @xmath37 is the radial pair distribution function @xcite .
this expression can also be derived from a purely dynamical description of colloidal motion , taking account of brownian motion and hydrodynamic interactions @xcite .
( [ eq : vhpluscorr ] ) provides an analogous expression for the osmotic pressure , for a system in which the solvent degrees of freedom are treated explicitly and on the same footing as those of the solute .
our simulations allow us to investigate in detail the microscopic forces which produce the imbalance in the density of solvent particles between the solution and solvent compartments .
figure [ fig : force ] shows the net force per particle , acting on the solvent particles , as a function of position @xmath38 , in a slab through the middle of the simulation box .
solvent particles close to the boundaries of the solution compartment are on average pushed out of the solution , towards the solvent compartment .
this net force arises from the fact that the total density inside the solution compartment is higher than that in the solvent compartment ( as shown in figure [ fig : density1]a ) thus , we expect solvent particles at the boundary to experience more collisions from the solution side than from the solvent side . )
acting on the solvent particles , as a function of position @xmath38 , in a slab taken through the middle of the simulation box , for solute concentration @xmath39 .
solvent particles close to the boundaries of the solution compartment tend to be pushed out of the solution .
, scaledwidth=40.0% ] why then is there no net flow of solvent out of the solution compartment in response to the net outward force ?
the answer lies in the solvent density profiles .
as shown in figure [ fig : density1]b , the density of _ solvent _ particles is higher in the solvent compartment than in the solution .
we expect this to create a diffusive flux of solvent particles into the solution , which counteracts the outward flux caused by the net outward force .
considering , rather simplistically , the `` hopping '' of solvent particles between layers next to the membrane on the solvent and solution sides , we expect that an individual particle on the solution side has a high probability of leaving the solution ( i.e. to hop from the solution to the solvent compartment ) , due to the outward pushing force , while an individual particle on the solvent side has only a low probability of entering the solution ( i.e. to hop from the solvent to the solution compartment ) . however , there are more solvent particles on the solvent side and consequently more attempts to hop into the solution than out of it .
thus the net fluxes of solvent particles inward and outward are equal and the solvent density imbalance is maintained .
this picture has much in common with the classic colloidal sedimentation equilibrium , in which an inhomogeneous colloidal density profile is maintained by a balance of a downward flux driven by the gravitational field and an upward diffusive flux .
this analogy between osmosis and sedimentation was also noted by soodak and iberall @xcite .
we note that these density profiles are also consistent with the statistical mechanical relation between the density and the potential of mean force @xcite .
however , the latter relation does not contain any dynamical information . using this simple `` hopping '' picture
, we can relate the solvent density imbalance to the osmotic pressure by simple mechanical arguments .
let us assume the solvent densities , @xmath25 and @xmath26 in the solution and solvent compartments respectively , are uniform , with a very sharp density change at the boundary .
we now define slabs of width @xmath40 ( of the order of the molecular size ) next to the boundary on the solvent and solution sides , and consider the hopping of solvent particles between these slabs .
solvent particles that hop from the solvent side into the solution experience an energy penalty @xmath41 , where the outward force @xmath42 arises from the fact that the total density is higher in the solution , and can be related to the osmotic pressure @xmath17 by @xmath43 , where @xmath44 is the slab area per particle . setting the fluxes of solvent particle hops equal in the inward and outward directions
, we find that @xmath45}= \rho_{\mathrm{v}}^{\mathrm{out}}\exp{\left[-ah \delta p /(k_bt)\right]}$ ] .
noting now that @xmath46 is the volume per particle , which is approximately the same on both sides of the membrane
i.e. that @xmath47 , leads to the result @xmath48 } .
\label{eq : hopping1}\ ] ] eq .
( [ eq : hopping1 ] ) is in fact a standard relation , which can be derived via thermodynamic arguments @xcite , by setting the solvent chemical potential equal across the membrane and making the assumption that @xmath49 .
thus our simple hopping model provides a clear mechanistic description which leads to the same predictions as the thermodynamic / statistical mechanical pictures .
the mechanistic description proposed here of course relies on the presence of repulsive interparticle interactions , which generate the `` outward pushing force '' . in a hypothetical case where the solute and solvent particles do not interact with each other ( `` ideal gas '' )
, this argument would not hold .
however in this case there would also be no osmotic density imbalance ; the solvent density would be equal on both sides of the membrane @xcite .
figure [ fig : dp - vs - conc ] ( inset ; triangles ) shows that eq .
( [ eq : hopping1 ] ) holds in our simulations , even for solute concentrations well beyond the limit of validity of the van t hoff relation .
the slight discrepancy between the prediction of eq .
( [ eq : hopping1 ] ) and the simulation results arises from the assumption @xmath50 made in our derivation
. for low solute concentrations , eq.([eq : hopping1 ] ) reduces to the van t hoff relation . noting that @xmath51 leads to : @xmath52 } \approx -\frac{\delta p } { k_bt \rho_{\mathrm{v}}^{\mathrm{out}}}. \label{eq : hopping2}\ ] ] the van t hoff relation is then recovered when we expand the logarithm to first order in @xmath53 ( which is small for low solute concentration ) : @xmath54 } \approx k_bt c_{\mathrm{s}}. \label{eq : hopping3}\ ] ] the simple picture presented above provides a mechanical description which is consistent with results obtained from thermodynamic arguments , and which is also consistent with the force profiles measured in our simulations . to show that this actually describes what is happening in our simulations
, we need to rule out more complex dynamical mechanisms .
a key feature of our simple picture is that it does not predict any correlations between the configuration of the solute particles and the hops of solvent particles across the membrane - unlike some other proposed mechanisms in which solvent particles are pulled across the membrane in the wake of nearby solute particles . to look for such effects in our simulations
, we therefore test whether correlations between solute and solvent motion influence the dynamics of solvent particles close to the membrane .
we first ask whether the microscopic flux of solvent particles into the solution compartment is influenced by the presence of nearby solute particles . at time @xmath55 , we focus on solvent particles which are situated less than 0.001 particle diameters will move in one timestep . ] outside the solution compartment . within the next timestep ,
some of these solvent particles cross into the solution compartment , while others do not .
figure [ fig : velocities]a shows the distribution of distances to the nearest solute particle , for those solvent particles which do cross in the next timestep ( black line ) and for those which do not ( red line ) .
these distributions are indistinguishable , suggesting that proximity of a solute particle does not make any difference to the chance that a solvent particle , located close to the boundary , will cross into the solution compartment .
one might also suppose that the _ velocity _ of nearby solute molecules might influence the microscopic flux of solvent particles
e.g. a solvent particle could cross into the solution in the wake of a nearby solute particle as it rebounds from the membrane .
figure [ fig : velocities]b shows the probability distribution for the velocity of the nearest solute particle ( perpendicular to the solution boundary ) , for solvent particles located close to the solution boundary , which do and do not subsequently cross into the solution .
once again , the two probability distributions are indistinguishable .
is defined as the angle between the instantaneous velocity of a solvent particle @xmath30 as it enters the solution compartment and the vector joining it to particle @xmath32 within the solution compartment .
( b ) : probability distribution @xmath56 when particle @xmath32 is defined to be the nearest solute particle ( black line ) and when particle @xmath32 is the nearest particle of either solute or solvent ( red line ) . in these simulations @xmath57 .
the boundary of the solution compartment is defined as described in appendix [ app : mop ] .
, scaledwidth=40.0% ] as a final test , we measured whether solvent crossings occur preferentially in the _ direction _ of the nearest solute particle .
if solute particles tend to `` suck '' solvent particles across the membrane , solvent crossings should be preferentially oriented towards the nearest solute particle .
we therefore measured the angle @xmath58 between the instantaneous velocity @xmath59 , of a solvent particle as it crossed into the solution compartment , and the vector @xmath60 , where @xmath61 and @xmath62 are the position of the solvent particle and the nearest solute , respectively ( see figure [ fig : angles]a ) .
figure [ fig : angles]b ( black line ) shows the probability distribution @xmath56 .
we then repeated the calculation , but this time allowing particle @xmath32 to be either solute or solvent ( within the solution compartment ) .
the resulting distribution is given by the red line in figure [ fig : angles]b .
again , the two distributions are indistinguishable , showing that solvent particles do not cross the membrane preferentially in the direction of nearby solute particles .
we therefore conclude that correlations between solute and solvent dynamics do not play a significant role in determining the microscopic solvent flux in our simulations .
this suggests that the simple dynamical picture described above is indeed sufficient to describe osmosis in our system .
in this paper , we have used a minimal model to investigate the dynamic mechanisms underlying the maintenance of equilibrium osmotic pressure and density gradients . in our system ,
solute and solvent particles interact identically , and there are no solvent - membrane interactions . for this system
, we are able to derive an expression ( eq.([eq : vhpluscorr ] ) ) for the osmotic pressure , in terms of the forces of interaction between solute - solute and solute - solvent pairs of particles ; this derivation does not assume any separation of scales between the solvent and solute particles .
analysis of the density and force profiles in our simulations leads us to propose a very simple picture , in which the density imbalance of solvent particles across the membrane is maintained by a balance between an outward force - driven flux ( due to the higher total density in the solution ) and an inward diffusive flux ( due to the lower solvent density in the solution ) .
we show that a simple calculation , based on the mechanics of hopping of solvent molecules across the membrane , can be used to derive the same relation between the solvent density gradient and the osmotic pressure ( eq.([eq : hopping1 ] ) ) as is obtained by standard thermodynamic arguments . for low solute concentrations
, this expression reduces to the van t hoff relation .
we do not detect any evidence of correlations between the solute particles and the solvent flux , leading us to believe that this simple `` hopping '' picture is sufficient to describe the dynamical basis of osmosis in our simulations . in previous work , various molecular mechanisms have been implicated in osmosis , including solute - membrane collisions , diffusion of solvent across the membrane , solute - solvent interactions and correlations between solvent and solute dynamics . in simulations of realistic osmotic systems , the dynamics of solvent molecules inside pores @xcite , interactions with the pore wall @xcite , interactions between solvent and solute molecules @xcite , and correlations between solvent flux and solute dynamics @xcite have all been shown to play important roles .
it is possible that different dynamical mechanisms are at play in different osmotic systems , all of these being consistent with the known , and universal , thermodynamic relations .
thus , an important topic for further work will be to understand how the conclusions of our work change for more realistic scenarios : e.g. when the membrane presents a barrier also to the solvent molecules ( via explicit pores or via a smooth potential barrier ) , when the solute and solvent do not interact identically , etc .
our aim in this work , however , was to focus on a minimal model system , for which we could characterize in detail the underlying molecular mechanisms .
for this system , we find that complex mechanisms are not required to understand the basic physics of osmosis .
we hope that this study will provide a basis for developing our understanding of osmosis in systems out of equilibrium , for which thermodynamic concepts are not available . recently
, much attention has focused on active particles such as motile bacteria , swimming `` janus '' colloids or colloidal `` chuckers '' @xcite .
these can act as solutes : an interesting question concerns how the osmotic pressure of a suspension of active colloids differs from that of the corresponding passive suspension .
moreover , active particles may themselves generate osmotic gradients , with interesting potential for motility and self - organisation .
the approach described here can easily be extended to investigate the dynamics of osmotic systems in which either the solute or the solvent particles are active . finally , we note that the minimal setup presented in this paper also provides a very general way to measure osmotic pressure in systems involving solvent and solutes , simply by partitioning a system into two compartments using an external potential permeable only to the solvent . indeed , a similar approach has already been used to fit intermolecular potentials to experimental data on osmotic pressure @xcite .
this kind of setup should facilitate osmotic pressure measurements in instrinsically non - equilibrium systems ( e.g. with self - propelled particles ) , where thermodynamic approaches are not feasible .
the authors thank daan frenkel for drawing this topic to our attention , leo lue for assistance with the calculations in appendix [ app : cs ] , mike cates , davide marenduzzo and aron yoffe for useful discussions and mike cates and patrick warren for helpful comments on the manuscript .
twl was supported by an epsrc dta studentship and rja by a royal society university research fellowship and by epsrc under grant ep / eo30173 .
this work has made use of the edinburgh compute and data facility , which is partially supported by the edikt initiative .
we measure the local pressure in the solution and solvent compartments using the method of planes @xcite . in this method
, one defines a plane within the region of interest , and monitors the flux @xmath63 of particle momentum normal to the plane ( due to particles crossing the plane ) , and the interparticle forces normal to the plane , for pairs of particles on opposite sides of the plane . for a plane at position @xmath64 , which extends right across the simulation box in the @xmath65 and @xmath66 directions ,
the pressure @xmath67 is given by @xmath68 f_{xij } , \label{eq : mop}\ ] ] where @xmath69 is the area of the plane , the sum is over all pairs of particles ( without double counting ) , @xmath70 and @xmath71 are the x - components of the particle position ( @xmath72 being the normal to the plane ) and @xmath73 is the normal component of the force between particles @xmath30 and @xmath32 .
the term in the square brackets ensures that only pairs of particles which are on opposite sides of the plane contribute to @xmath67 .
expression ( [ eq : mop ] ) describes the instantaneous pressure , which has @xmath74-function peaks whenever a particle crosses the plane ; this expression is time - averaged to provide a meaningful pressure measurement . and @xmath66 directions .
solid unidirectional arrows denote contributions to the kinetic part of eq.([eq : mop ] ) ( @xmath63 ) ; solid bidirectional arrows denote contributions to the interaction part of eq.([eq : mop ] ) . all interactions in which the line connecting the two particles crosses the plane
are included ( e.g. the bottom pair of particles ) ; however the interaction between the top pair of particles ( dotted arrow ) is excluded .
( b ) : pressure - density relation of a `` gas '' of solute particles , confined in the solution compartment in the absence of solvent ( closed circles ) , compared with that of a periodic box of equivalent particles ( open circles ) .
the pressure is in units of @xmath75 ; the density is in units of @xmath16 . for the closed circles ,
the density of the solute gas is defined using an effective boundary of the solution compartment positioned @xmath76 inside the locus of divergence of the the confining potential.,scaledwidth=40.0% ] in our simulations , we modify this method in order to use planes with finite extent in the @xmath65 and @xmath66 directions . in this case
, the sum in equation [ eq : mop ] should include only those pairs of particles for which the line connecting the particle positions crosses the plane ( see figure [ fig : mopdiag]a ) .
we have verified that this method produces results in good agreement with global pressure measurements ( for homogeneous systems ) and with other methods for measuring the local pressure , even when the plane is very small , of the order of a few particle diameters . in our system , since there are no solvent - membrane interactions , the osmotic pressure is also given by the magnitude of the forces of confinement on the solute particles , per unit area of the confining boundary ( as noted by luo and roux @xcite ) .
this provides an alternative way to measure the osmotic pressure , with results in good agreement with various other methods for determining the local pressure @xcite . to define the concentration of solute particles in our simulations , we need to define an effective boundary for the solution compartment . to do this , we match the pressure - density relation of a `` gas '' of solute particles , confined in the solution compartment in the absence of solvent ( shown in figure [ fig : mopdiag]b , closed circles ) , with that of a periodic box of equivalent particles ( figure [ fig : mopdiag]b , open circles )
this results in an effective boundary of the solution compartment which lies @xmath77 inside the locus of divergence of the the confining potential .
the error bars on the position of this boundary were computed by determining the range of boundary positions over which the simulated pressure - density curve for the confined solute `` gas '' can be said to match both the hard - sphere carnahan - starling prediction and simulation results for an unconfined ( periodic ) gas of wca particles , with 95% confidence ( evaluated using a t - test ) .
in figure [ fig : density2 ] , we compare our simulation results for the solvent density imbalance between solution and solvent compartments , to thermodynamic predictions for a system of hard spheres , obtained using the carnahan - starling expression for the free energy of a system of hard spheres @xcite .
the solvent compartment , of volume @xmath78 , contains @xmath79 solvent particles , at packing fraction @xmath80 ( this is related to the density @xmath81 by @xmath82 ) .
the free energy is then @xmath83\\\nonumber & = & \frac{6\eta_{\mathrm{v}}^{\mathrm{out } } v^{\mathrm{out}}}{\pi\sigma^3}\left[\log\left(\frac{6 \eta_{\mathrm{v}}^{\mathrm{out } } \lambda^{3}}{\pi \sigma^3}\right)-1 + \eta_{\mathrm{v}}^{\mathrm{out}}\frac{(4 - 3\eta_{\mathrm{v}}^{\mathrm{out}})}{(1-\eta_{\mathrm{v}}^{\mathrm{out}})^{2}}\right ] \label{eq : hsfreeenergy}\end{aligned}\ ] ] where @xmath84 is the de broglie thermal wavelength .
the chemical potential is given by @xmath85 : @xmath86 \label{eq : musolvent1}\end{aligned}\ ] ] the solution compartment , of volume @xmath87 , contains a mixture of solute particles , at packing fraction @xmath88 , and solvent particles , at packing fraction @xmath89 .
its free energy is @xmath90 + n_{\mathrm{v}}^{\mathrm{in}}\left[\log\left(\rho_{\mathrm{v}}^{\mathrm{in } } \lambda^{3}\right)-1 \right]\\\nonumber & + & \frac{6v^{\mathrm{in}}(\eta_{\mathrm{s}}+\eta_{\mathrm{v}}^{{\mathrm{in}}})^2 } { \pi\sigma^3}\left[\frac{4 - 3(\eta_{\mathrm{s}}+\eta_{\mathrm{v}}^{{\mathrm{in}}})}{1-(\eta_{\mathrm{s}}+\eta_{\mathrm{v}}^{{\mathrm{in}}})^{2}}\right ] \label{eq : hsfreeenergy2}\end{aligned}\ ] ] the chemical potential @xmath91 : @xmath92 .
\label{eq : musolvent2}\end{aligned}\ ] ] setting @xmath93 , we arrive at the following relation : @xmath94\\ & = & \eta_{\mathrm{v}}^{{\mathrm{out}}}\left[\frac{8 - 9\eta_{\mathrm{v}}^{{\mathrm{out}}}+3(\eta_{\mathrm{v}}^{{\mathrm{out}}})^{2}}{(1-\eta_{\mathrm{v}}^{{\mathrm{out}}})^{3}}\right ] .
\label{eq : densitydiff}\end{aligned}\ ] ] conservation of solvent particle number implies that @xmath95 where @xmath96 is the total number of solvent particles .
( [ eq : densitydiff ] ) and ( [ eq : cons ] ) are solved numerically to obtain @xmath89 and @xmath80 as functions of the solute packing fraction @xmath88 , as plotted in figure [ fig : density2 ] .
( [ eq : vhpluscorr ] ) follows from the dynamical equations of motion for the solute particles , under the conditions that the system is in steady state and the solute particles are confined in the solution compartment .
the first of these conditions implies that @xmath97 ( where the sum is over all solute particles and @xmath98 and @xmath99 are the position and momentum of particle @xmath30 respectively ) . expanding the time derivative and using newton s equations of motion leads to : where @xmath101 is the total force acting on solute particle @xmath30 , and @xmath102 is its mass .
eq.([eq : clausius ] ) constitutes a clausius virial relation for the solute particles @xcite .
splitting the force @xmath101 into the contributions due to interactions with the confining potential , solute - solute interactions and solute - solvent interactions , and using the equipartition theorem , we arrive at : where @xmath31 is the force exerted on particle @xmath30 by particle @xmath32 and @xmath104 is the confining force on particle @xmath30 .
the first sum is over solute - solute pairs ( without double counting ) , the second sum is over solute - solvent pairs , and the final sum is over interactions between solute particles and the `` membrane '' .
focusing on the final term , we note that , to a good approximation , solute particles only feel the confining force when they are very close to the boundary of the solution compartment . the final term can then be approximated by an integral over a narrow shell at the boundary of the solution compartment .
taking into account also the fact that the osmotic pressure is given by the average normal force per unit area on the membrane , which in our system is equivalent to the magnitude of the confining force on the solute particles , per unit area , leads to @xcite where @xmath17 is the osmotic pressure and @xmath87 is the volume of the solution compartment .
substituting eq.([eq : os ] ) into eq.([eq : solutevirial ] ) leads directly to eq.([eq : vhpluscorr ] ) in the main text . | osmosis is one of the most important physical phenomena in living and soft matter systems .
while the thermodynamics of osmosis is well understood , the underlying microscopic dynamical mechanisms remain the subject of discussion .
unraveling these mechanisms is a prerequisite for understanding osmosis in non - equilibrium systems . here , we investigate the microscopic basis of osmosis , in a system at equilibrium , using molecular dynamics simulations of a minimal model in which repulsive solute and solvent particles differ only in their interactions with an external potential . for this system , we can derive a simple virial - like relation for the osmotic pressure .
our simulations support an intuitive picture in which the solvent concentration gradient , at osmotic equilibrium , arises from the balance between an outward force , caused by the increased total density in the solution , and an inward diffusive flux caused by the decreased solvent density in the solution .
while more complex effects may occur in other osmotic systems , our results suggest that they are not required for a minimal picture of the dynamic mechanisms underlying osmosis . |
quenched randomness effects are very important in modeling real materials , since consideration of these effects simulates a realistic model of real materials .
real materials have some uncontrollable defects and these defects can be modeled by introducing site dilution ( randomly distributed non magnetic atoms ) , bond dilution ( randomly broken bonds between the magnetic atoms ) or both of them into the related model .
it is a well known fact that , heisenberg model produces more realistic results than the ising model , in order to explain the magnetic properties of real materials .
thus , it is important to work on the heisenberg model with these quenched randomness effects .
these quenched randomness effects produce different behaviors in the magnetic properties of the model , e.g. different phase transition characteristics from the pure model ( i.e. the model without any quenched randomness effects ) .
heisenberg model with quenched randomness effects has been studied widely by a variety of methods , such as spin-1/2 ( s-1/2 ) isotropic heisenberg model with bond dilution on 2d lattices with monte carlo ( mc ) simulation @xcite , anisotropic model ( by means of the xxz model ) on 2d lattices with real space renormalization group ( rsrg ) technique @xcite . on the other hand ,
random bond distributed systems in which spin glass phases originate have been studied , e.g. discrete distribution on s-1/2 heisenberg model with pair approximation @xcite , density matrix product approximation @xcite , gaussian distribution with imaginary time grassmann field theory @xcite and s-1 heisenberg model with discretely distributed random bonds with exact diagonalization method@xcite . besides , both site and bond diluted systems have been studied with quantum mc on 2d lattices @xcite and s-1/2 heisenberg model with site - bond correlated dilution ( which covers uncorrelated site dilution as a limit ) on 2d and 3d lattices with rsrg @xcite and mc @xcite , and also using a variational principle for the free energy @xcite .
all of these work are related to the isotropic heisenberg model or xxz model .
the aim of this work is determine the effect of the bond dilution on the phase diagrams and the thermodynamic properties of the anisotropic quantum heisenberg model .
by anisotropy , we do not restrict ourselves with xxz model .
namely , we want to determine the effect of the anisotropy in the exchange interaction on the phase transition characteristics not only by means of xxz model .
the method is effective field theory ( eft ) with two spin cluster approximation@xcite .
eft approximation can provide results that are superior to those obtained within the traditional mean field approximation ( mfa ) , due to the consideration of self spin correlations which are omitted in the mfa .
eft for a typical magnetic system starts by constructing a finite cluster of spins which represents the system .
callen - suzuki spin identities @xcite are the starting point of the eft for the one spin clusters .
if one expands these identities with differential operator technique@xcite , multi spin correlations appear , and in order to avoid from the mathematical difficulties , these multi spin correlations are often neglected by using decoupling approximation ( da ) @xcite . working with larger finite clusters will give more accurate results .
callen - suzuki identities have been generalized to two spin clusters in ref .
@xcite ( namely eft-2 formulation ) .
this eft-2 formulation has been successfully applied to a variety of systems , such as quantum s-1/2 heisenberg ferromagnet @xcite and antiferromagnet @xcite systems , classical n - vector model @xcite , and spin-1 heisenberg ferromagnet @xcite .
the paper is organized as follows : in sec .
[ formulation ] , we briefly present the model and formulation . the results and discussions are presented in sec .
[ results ] , and finally sec .
[ conclusion ] contains our conclusions .
we consider a lattice which consists of @xmath0 identical spins ( s-@xmath1 ) such that each of the spins has @xmath2 nearest neighbors .
the hamiltonian of the system is given by @xmath3 where @xmath4 and @xmath5 denote the pauli spin operators at a site @xmath6 .
@xmath7 and @xmath8 stand for the components of the exchange interaction @xmath9 ( in other words anisotropy in the exchange interaction ) between the nearest neighbor spins @xmath6 and @xmath10 .
the sum is carried over the nearest neighbors of the lattice .
the bonds between the spins @xmath6 and @xmath10 are randomly distributed via @xmath11 in the bond dilution problem . the distribution given by eq .
( [ denk2 ] ) distribute bonds randomly between lattice sites as @xmath12 percentage of bonds are closed and remaining @xmath13 percentage of bonds are open , i.e. @xmath12 is the concentration of closed bonds in the lattice . here
@xmath14 stands for the delta function and @xmath12 is a real number which is defined within the range of @xmath15 . the distribution given by eq .
( [ denk2 ] ) reduces to the system with homogenously distributed bonds ( i.e. pure system ) for @xmath16 .
we use the two spin cluster approximation as an eft formulation , namely eft-2 formulation@xcite . in this approximation , we choose two spins ( namely @xmath17 and @xmath18 ) and treat the interactions in this two spin cluster exactly . in order to avoid some mathematical difficulties we replace the perimeter spins of the two spin cluster by ising spins ( axial approximation ) @xcite . with the procedure defined in ref .
@xcite , we get an expression for the magnetization per spin as @xmath19 where @xmath20 , @xmath21 is boltzmann constant and @xmath22 is the temperature . the inner average bracket in eq . ( [ denk3 ] ) ( which has no subscript ) stands for thermal average and the outer one ( which has subscript r ) is for the configurational averaging which is necessary for including the effect of the random bond distribution . the parameters in eq .
( [ denk3 ] ) are given by @xmath23^{1/2}\\ y_0^\prime&=&\left[{\left(j^{(12)}_x+j^{(12)}_y\right)}^2+h_{-}^{\prime 2}\right]^{1/2}\\ \end{array}\ ] ] @xmath24 @xmath25 where @xmath26 stands for the @xmath2 component of the nearest neighbor of the spin @xmath17 while @xmath27 stands for the @xmath2 component of the nearest neighbor of the spin @xmath18 .
the sums in the eq .
( [ denk6 ] ) are over the nearest neighbor sites of the sites labeled @xmath28 and @xmath29 , respectively .
the configurational averages can be calculated via integration of the expression by using eq .
( [ denk2 ] ) , over the all bonds of the treated cluster . @xmath30 where product is taken over the nearest neighbors of the sites @xmath28 and @xmath29 .
if we perform the integration concerning the bond between the sites @xmath28 and @xmath29 then we get the following expression , @xmath31 @xmath32\right\rangle\ ] ] where @xmath33^{1/2}\\ y_0^{\prime\prime}&=&\left[{\left(j_x+j_y\right)}^2+h_{-}^{\prime 2}\right]^{1/2}.\\ \end{array}\ ] ] eq .
( [ denk8 ] ) can be written in terms of the differential operators using differential operator technique @xcite and it is found as @xmath34 where @xmath35 and @xmath36^{1/2}\ ] ] @xmath37^{1/2}\ ] ] @xmath38 in eq .
( [ denk10 ] ) , the parameters @xmath39 and @xmath40 are the usual differential operators in the differential operator technique .
differential operators act on an arbitrary function via @xmath41 with any constant @xmath42 and @xmath43 .
now let us assume that each of @xmath17 and @xmath18 has number of @xmath44 distinct nearest neighbors and both of them have @xmath45 common nearest neighbors .
this means that , in eq .
( [ denk6 ] ) , the two sums have number of @xmath45 common terms . if we take the integration in eq .
( [ denk10 ] ) with the help of eqs .
( [ denk2 ] ) and ( [ denk6 ] ) and using da @xcite we get an expression , @xmath46^{z_0 } \left[a_{y}+m b_{y}\right]^{z_0 } \left[a_{xy}+m b_{xy}\right]^{z_1}\right\rangle } f{\left(x , y\right)}|_{x=0,y=0}\ ] ] for the magnetization .
the coefficients are defined by @xmath47}+(1-c)&\quad & b_{xy}=c\sinh{\left[j_z{\left(\nabla_x+\nabla_y\right)}\right]}.\\ \end{array}\ ] ] with the help of the binomial expansion , eq .
( [ denk14 ] ) can be written as @xmath48 where the coefficients are @xmath49 and these coefficients can be calculated by using the definitions given in eqs .
( [ denk11]),([denk13 ] ) and ( [ denk15 ] ) .
let us write eq .
( [ denk16 ] ) in more familiar form as @xmath50 @xmath51 where @xmath52 is the kronecker delta . for a given set of hamiltonian parameters ( @xmath53 ) , temperature and bond distribution parameter ( @xmath12 ) , we can determine the coefficients from eq .
( [ denk17 ] ) and we can obtain a non linear equation from eq .
( [ denk18 ] ) . by solving this equation
, we can get the magnetization ( @xmath54 ) for a given set of parameters and temperature .
since the magnetization is close to zero in the vicinity of the critical point then we can obtain a linear equation by linearizing the equation given in eq .
( [ denk18 ] ) which allows us to determine the critical temperature .
since we have not calculated the free energy in this approximation , we can locate only second order transitions from the condition given as @xmath55the tricritical point at which second and first order transition lines meet can be determined from the condition @xmath56
let us scale the exchange interaction components with the unit of energy @xmath57 as @xmath58 where @xmath59 .
let us choose @xmath60 , then @xmath61 can be used as the measure of the anisotropy in the exchange interaction .
it can be seen from the formulation of the problem ( e.g. see eqs .
( [ denk11])-([denk15 ] ) ) that the transformation @xmath62 does not change the expression of the order parameter , regardless of the bond distribution .
hence , we can say that aforementioned transformation does not affect the formulation .
therefore , let us fix @xmath63 and concentrate only on the varying @xmath64 values .
our investigation will be focused on selected several 2d and 3d lattices in isotropic case and on a honeycomb lattice in anisotropic case with the ferromagnetic exchange interaction , i.e. @xmath65 .
it can be seen from the hamiltonian of the system which is given by eq .
( [ denk1 ] ) that for @xmath66 , @xmath67 the system reduces to a model which has ising type symmetry and @xmath68 or @xmath69 reduces the system to the system that have @xmath70 type symmetry . in this case , all components of the exchange interaction are equal to each other , i.e. @xmath71 in eq .
( [ denk22 ] ) . in order to represent the behavior of the critical temperature ( @xmath72 ) as a function of the bond concentration ( @xmath12 ) for the isotropic quantum heisenberg model
, we depict the phase diagrams for three different lattices in ( @xmath73 ) plane in fig .
( [ sek1])(a ) . as we can see from fig .
( [ sek1])(a ) that the critical temperature of the system gradually decreases when the bond concentration decreases and takes the value of zero at a specific value of bond concentration , namely at the bond percolation threshold ( @xmath74 ) which depends on the lattice geometry . at this point
, we have to mention that bond percolation threshold value may not correspond to the value that makes the critical temperature zero .
if the phase diagram of the system exhibits a reentrant behavior in the ( @xmath73 ) plane then the bond percolation threshold value corresponds to the @xmath12 value that terminates the phase diagram , i.e. for the value that provide @xmath75 system can not reach the ordered phase any more
. in the isotropic case of the quantum heisenberg model , the whole phase diagrams in the ( @xmath73 ) plane do not exhibit reentrant behavior .
the numerical values of bond percolation threshold values obtained within the present work are @xmath76 for a honeycomb lattice , @xmath77 for a square lattice , and @xmath78 for a simple cubic lattice .
we can also see from fig .
( [ sek1])(a ) that at a certain @xmath12 , simple cubic lattice has highest critical temperature value which is due to the large number of the nearest neighbor spins in the lattice .
besides , in fig .
( [ sek1])(b ) we can see the variation of the ground state magnetization ( @xmath79 ) with the bond concentration for three lattice geometries .
these values of magnetizations calculated at the temperature @xmath80 can be treated as ground states since the energy comes from the thermal fluctuations are negligible in comparison with the energy comes from the spin - spin interaction . as we can see from fig .
( [ sek1])(b ) that , ground state magnetization starts from the value of @xmath81 and , stays in this value for a while when the bond concentration decreases .
after then , it starts to fall and reach the zero at a bond percolation threshold value .
it is a known fact that for a certain set of hamiltonian parameters the critical temperature of the pure system with ising type symmetry is greater than the system with xy type symmetry and this value is greater than that of the system with heisenberg type symmetry@xcite .
from this fact we expect that , the relation between the bond percolation threshold values of three types of symmetries to be @xmath82 . here , the letters in the parenthesis depict the symmetry type of the hamiltonian .
namely , @xmath83 stands for heisenberg type symmetry ( with the anisotropy values @xmath84 ) , @xmath70 stands for xy type symmetry ( with the anisotropy values @xmath85 ) , and @xmath86 stands for ising type symmetry ( with the anisotropy values @xmath87 ) .
but the formulation gives this relation as @xmath88 , as seen in table ( 1 ) .
we will touch upon this point in the next subsection .
the bond percolation threshold values of the different lattices for different hamiltonian symmetries can be seen in table ( 1 ) .
[ table1 ] lllllllll lattice & @xmath44 & @xmath45 & @xmath89 ( i)&@xmath89 ( xy)&@xmath89 ( h ) + honeycomb & 2 & 0 & 0.5707&0.6448&0.5903 + kagome & 2 & 1 & 0.4655&0.5092&0.4709 + square & 3 & 0 & 0.4295 & 0.4699&0.4427 + triangular & 3 & 2 & 0.3102 & 0.3289 & 0.3127 + simple cubic & 5 & 0 & 0.2903 & 0.3085 & 0.2974 + body centered cubic & 7 & 0 & 0.2200 & 0.2305 & 0.2245 + face centered cubic & 7 & 4 & 0.1537 & 0.1584 & 0.1546 + + the effect of the bond concentration on the behavior of the magnetization with temperature can be seen in figs .
( [ sek2])(a ) for a honeycomb lattice and ( [ sek2])(b ) for a simple cubic lattice . as seen in fig.([sek2 ] ) that , decreasing @xmath12 causes a decline in the critical temperature value , also after some value of it , it causes decline in the ground state magnetization .
ground state magnetization starts to decrease at a lower value of @xmath12 for the simple cubic lattice ( in comparison with the honeycomb lattice ) , as seen in fig .
( [ sek2 ] ) in compatible with fig .
( [ sek1])(b ) ( e.g. all curves of @xmath12 value that provide @xmath90 for the simple cubic lattice have ground state magnetization @xmath81 while for the honeycomb lattice the same ground state magnetization value has been observed only for @xmath91 values ) .
again , this is due to the excess number of nearest neighbor spins of the simple cubic lattice in comparison with those of the honeycomb lattice .
now , let us see the effect of the anisotropy in the exchange interaction on the results obtained for the isotropic case in the preceding subsection . to do this , we start with the phase diagrams corresponding fig .
( [ sek1 ] ) ( a ) for the isotropic case .
the phase diagrams for a honeycomb lattice in ( @xmath73 ) plane can be seen in fig .
( [ sek3 ] ) for different anisotropy values . in fig .
( [ sek3])(a ) , we can see the phase diagrams of the system with ising type symmetry ( the curve labeled by a ) and xy type symmetry ( curves labeled by b , c , d ) . as we can see from the curves labeled by b , c , d in fig .
( [ sek3])(a ) , as the amount of anisotropy increases then the ferromagnetic region gets narrower in the ( @xmath73 ) plane for the xy type model and after a certain value of @xmath12 , phase diagrams disappears .
this value depends on the lattice geometry .
bond percolation threshold values of the xy type system gradually decrease when the anisotropy increases .
there is not any reentrant behavior observed in the phase diagrams of the xy type system . in figs .
( [ sek3 ] ) ( b ) and ( c ) , transition from the curves labeled by a to the other curves corresponds to the transition from the system with xy type symmetry to the system with h type symmetry .
when we take a look at fig .
( [ sek3 ] ) ( b ) , we see that increasing anisotropy effects the phase diagrams in the ( @xmath73 ) plane in a more complicated way in comparison with fig .
( [ sek3 ] ) ( a ) .
firstly , as the anisotropy increases then the bond percolation thresholds decrease for a while then tend to increase .
we can clearly see here , although the critical temperature of the pure system for the xy type symmetry ( the curve labeled a in fig .
( [ sek3])(b ) ) is greater than that of the h type symmetry ( curve labeled c in fig .
( [ sek3])(b ) ) , this does not imply that the bond percolation threshold value of the xy type symmetric model to be lower than the h type symmetric model . this behavior of the curves labeled a and c seen in fig .
( [ sek3])(b ) verifies the results in table ( 1 ) .
secondly , increasing anisotropy can induce a reentrant behavior in the phase diagrams ( e.g. see the curve labeled by b in fig .
( [ sek3 ] ) ( b ) ) . besides
, we can see from fig .
( [ sek3 ] ) ( c ) that , increasing anisotropy in the exchange interaction can produce phase diagrams that behave qualitatively different from the others . for instance , for the curve labeled by e in fig .
( [ sek3 ] ) ( c ) , decreasing anisotropy first causes a sharp decrease in the critical temperature then the critical temperature stays almost constant before it depresses to zero . in this figure , the curves labeled by b and d also exhibit reentrant behavior .
one other interesting behavior can be seen in fig .
( [ sek3 ] ) ( d ) . for @xmath92 , with low anisotropies ( i.e. small values of @xmath64 ) , there are no phase diagrams .
the system can not exhibit an ordered phase . when the value of @xmath64 increases then the phase diagrams appear and then disappear again . in conclusion
, we can say that while the transition from the @xmath86 type symmetry to the @xmath70 type symmetry ( fig .
( [ sek3 ] ) ( a ) ) rising anisotropy in the exchange interaction can not induce reentrant behavior in the phase diagrams which are in the @xmath93 plane .
but , if increasing anisotropy causes a transition from the @xmath70 type symmetry to the @xmath83 type symmetry then it can induce a reentrant behavior in the phase diagrams in the @xmath93 plane , ( fig .
( [ sek3 ] ) ( b ) and ( c ) ) .
in order to concentrate on the evolution of the bond percolation threshold value with anisotropy in the exchange interaction , let us investigate the variation of @xmath74 value with @xmath64 for some selected values of @xmath63 .
these curves can be seen in fig .
( [ sek4])(a ) . the curve labeled by a corresponds to the xy type system . as mentioned in the comments of fig .
( [ sek3 ] ) , for the xy type system , with increasing anisotropy , the bond percolation threshold value smoothly increases to the value @xmath94 while for the system with h symmetry , the bond percolation threshold value first decreases with increasing anisotropy then increases and reaches to the value of @xmath94 ( see the curves labeled by b , c , d in fig .
( [ sek4])(a ) ) .
when @xmath63 increases , the @xmath74 value corresponding to @xmath95 increases and after a specific value of @xmath96 the @xmath74 value for @xmath95 becomes @xmath94 .
this value is nothing but the value of the curve labeled by a intersects the @xmath97 , due to the symmetry of the formulation as mentioned at the beginning of this section .
this is just the value that makes the critical temperature of the pure system ( i.e. @xmath98 ) equal to zero with @xmath70 type symmetry .
we can numerically determine the similar borders of the model with @xmath83 type symmetry . as seen in fig .
( [ sek4 ] ) ( a ) , curves intersect the @xmath97 line at a specific value of @xmath64 which depends on the value of @xmath63 .
the curves labeled a , b and c intersect the @xmath97 line only by one point while , the curve labeled d by two point .
let denote these two intersection points as @xmath99 and @xmath100 which provide @xmath101 .
in other words @xmath100 denotes the value of @xmath64 which is related @xmath63 curve intersects to the @xmath102 line on the right side and @xmath99 denotes the value of @xmath64 which related @xmath63 curve intersects to the @xmath102 line on the left side , if present .
these two special values in the @xmath103 plane can be seen in fig .
( [ sek4 ] ) ( b ) . for any point which is outside of the region confined by this curve and @xmath61 axes
, we can say that the system can not exhibit an ordered phase for a honeycomb lattice .
this border consists of three segments : @xmath104 line , @xmath105 line and the remaining curve .
the region that the system can exhibit an ordered phase becomes expanded when @xmath63 increases from @xmath106 to @xmath94 , after then that region starts to shrink .
we can say from the symmetry properties of the formulation that , this border curve is symmetric about the @xmath107 line . we can obtain similar borders for the system which has bond concentration values different from @xmath108 .
these curves can be seen in fig .
( [ sek4 ] ) ( c ) .
again we can say with fig .
( [ sek4 ] ) ( c ) that , for any point which is outside of the region confined by the related @xmath12 curve and @xmath61 axes , the system with @xmath109 percentage open bonds , can not exhibit an ordered phase for a honeycomb lattice . finally , we want to investigate the effect of the bond dilution on the ground state magnetizations of the system . in the isotropic case ,
this effect was as decreasing @xmath12 gradually decreases the ground state magnetizations ( fig .
( [ sek1 ] ) ( b ) ) .
but in anisotropic case , this situation is a little complicated as seen in fig .
( [ sek5 ] ) , which is the variation of the ground state magnetization ( @xmath79 ) with anisotropy ( @xmath64 ) and bond concentration ( @xmath12 ) for some selected values of @xmath63 .
( [ sek5 ] ) ( a ) is just concerning the model with @xmath70 type symmetry .
there is not any complicated behavior as in the isotropic model . for a certain @xmath12
, the ground state magnetization regularly decreases with increasing @xmath64 .
but the situation is more complicated in figs .
( [ sek5 ] ) ( b ) and ( c ) .
these are the surfaces regarding the model with @xmath83 type symmetry and all the reentrant behaviors in the phase diagrams of the system ( e.g. curve labeled by @xmath110 in fig .
( [ sek3 ] ) ( b ) and ( c ) ) shows itself in the behavior of the ground state magnetizations . beside these , @xmath111 surface is not smooth while @xmath112 surface is smooth i.e. , for @xmath112 with any value of @xmath12 , increasing @xmath64 changes the ground state magnetization continuously but this is not the case for @xmath111 .
increasing @xmath63 value also shrinks this surface as expected .
( [ sek5])(d ) has different characteristics than the others ; the ground state magnetization surface can not reach the small valued @xmath64 regions , as well as the higher valued @xmath64 regions .
this fact is compatible with the curve labeled by d in fig .
( [ sek4 ] ) ( a ) .
the effect of the bond dilution on the phase diagrams of the isotropic and anisotropic quantum heisenberg model has been investigated in detail .
first in the isotropic model , effect of the bond dilution process on the critical temperatures have been given for honeycomb , square and simple cubic lattices .
there is not any reentrant behavior observed in that phase diagrams plotted in ( @xmath73 ) plane for the isotropic heisenberg model .
also , the variation of the ground state magnetization with bond concentration ( @xmath12 ) and the evolution of the variation of the magnetization with temperature curves with @xmath12 have been presented .
the bond percolation threshold values which is the value of @xmath12 that terminates the phase diagrams in the @xmath93 plane have been given for several 2d and 3d lattices with the hamiltonians that have different symmetries .
in addition , the same investigation has been performed for the anisotropic model .
anisotropic model has been treated within three different regions .
namely , the hamiltonian that have ising type symmetry @xmath87 , xy type symmetry @xmath113 and heisenberg type symmetry @xmath114 .
all these models that have different symmetry properties have been investigated in detail : the effect of the bond dilution process on both the ground state magnetizations and critical temperatures have been determined via the phase diagrams in the @xmath93 plane . beside these , the bond percolation threshold values ( @xmath74 ) have been determined for the anisotropic model and depicted in the @xmath115 plane for selected values of @xmath63 . for the xy type symmetric model , decreasing bond concentration results in gradually increase of the bond percolation threshold value .
there is not any reentrant behavior observed within the model that have xy type symmetry on the phase diagrams in the @xmath93 plane . on the other hand , in the model that have heisenberg type symmetry , this situation is more complicated
. increasing anisotropy does not results in a gradually increase of the @xmath74 value as in the model that have xy type symmetry .
however , increasing anisotropy first causes a decrease in the @xmath74 value then an increase in it before it reaches the value of @xmath94 . beside this fact ,
increasing anisotropy also give rise to the reentrant behavior in the phase diagrams in the @xmath93 plane .
all of these reentrant behaviors are found to be of the second order . | effects of the bond dilution on the critical temperatures , phase diagrams and the magnetization behaviors of the isotropic and anisotropic quantum heisenberg model have been investigated in detail . for the isotropic case ,
bond percolation threshold values have been determined for several numbers of two ( 2d ) and three ( 3d ) dimensional lattices . in order to investigate the effect of the anisotropy in the exchange interaction on the results obtained for the isotropic model , a detailed investigation has been made on a honeycomb lattice .
some interesting results , such as second order reentrant phenomena in the phase diagrams have been found .
keywords : * quantum anisotropic heisenberg model ; bond dilution ; bond percolation threshold * |
the quantum summation ( * qs * ) algorithm ( also known as the amplitude estimation algorithm ) of brassard , hyer , mosca and tapp , see @xcite , computes an approximation to the arithmetic mean of all values of a boolean function defined on a set of @xmath12 elements .
information regarding the boolean function is supplied by quantum queries .
the quantum queries play a role similar to the use of function values in the worst case and randomized settings .
suppose that we use @xmath1 quantum queries .
obviously , the only case of interest is when @xmath6 is much smaller than @xmath0 .
it was proven in @xcite that the error of the * qs * algorithm is at most @xmath13 nayak and wu , see @xcite , showed that for any @xmath14 $ ] the error of any quantum algorithm that uses no more than @xmath1 quantum queries must be proportional to @xmath11 with probability @xmath15 .
therefore , the * qs * algorithm enjoys the smallest possible error modulo a factor multiplying @xmath11 .
the minimal error estimate of order @xmath11 in the quantum setting should be compared to the minimal error estimates in the worst case and randomized settings of algorithms using @xmath1 function values .
it is known , see @xcite , that in the worst case setting , the error bound is roughly @xmath16 .
this means that as long as @xmath6 is much less than @xmath0 the error is almost @xmath17 , and is therefore of order @xmath6 times larger than in the quantum setting . in the randomized
setting , the classical monte carlo is almost optimal , and the error bound is roughly @xmath18 , see again @xcite .
hence , it is of order @xmath19 larger than in the quantum setting .
the * qs * algorithm has many applications . in particular , it can be used for approximation of the arithmetic mean of a real function , which is the basic step for approximation of many continuous problems such as multivariate integration , multivariate approximation and path integration , see @xcite .
since the * qs * algorithm has so many applications , it seems reasonable to check whether the estimate ( [ 111 ] ) is sharp and how the error decreases if we lower the probability @xmath20 to @xmath21 .
it also seems reasonable to study the error of the * qs * algorithm in various settings .
the estimate ( [ 111 ] ) corresponds to the worst - probabilistic setting , which is most frequently used in the quantum setting .
the essence of this setting is that it holds for all boolean functions .
it is also interesting to study the average performance of the * qs * algorithm with respect to some measure on boolean functions .
this is the average - probabilistic setting . in the worst - average and average - average settings , we study the worst or average performance with respect to boolean functions and the average performance with respect to all outcomes of a quantum algorithm . we add in passing that the worst - average setting is usually used for the study of the classical monte carlo algorithm .
sharp error bounds in the worst- and average - probabilistic settings are addressed in this paper whereas the worst- and average - average settings will be studied in a future paper .
we study error bounds with probabilities @xmath22 $ ] .
if we want to obtain error bounds with higher probability , it is known that it is enough to run the * qs * algorithm several times and take the median as the final result , see e.g. , @xcite . in the worst - probabilistic setting
, we show that ( [ 111 ] ) can be slightly improved .
namely , the error of the * qs * algorithm is at most @xmath23 furthermore , for large @xmath6 and @xmath24 we prove that the last estimate is sharp .
more generally , for @xmath25 $ ] we prove that the error of the * qs * algorithm is at most @xmath26 where @xmath27 is the inverse of the function @xmath28 .
we prove that the last estimate is sharp for large @xmath6 and @xmath24 .
we have @xmath29 $ ] and it is well approximated by @xmath30 .
in particular , for the most commonly used values of @xmath15 we have @xmath31 in the average - probabilistic setting , we consider two measures on the set of boolean functions .
the first measure is uniform on boolean functions , while the second measure is uniform on arithmetic means of boolean functions .
the results for these two measures are quite different .
the mean element of the arithmetic means is @xmath17 for both measures .
however , the first moment is of order @xmath32 for the first measure , and about @xmath33 for the second .
the first moment is exactly equal to the error of the constant algorithm that always outputs @xmath17 .
this explains why we can obtain the error of order @xmath32 without any quantum queries for the first measure .
this provides the motivation for us to check whether the error of the * qs * algorithm enjoys a similar property .
it turns out that this is indeed the case iff @xmath6 is divisible by @xmath9 .
that is , for @xmath6 divisible by @xmath9 , the average - probabilistic error of the * qs * algorithm is of order @xmath34 , and if @xmath6 is not divisible by @xmath9 , then the error is of order @xmath11 . for the second measure ,
since the first moment is not small , the average - probabilistic error of the * qs * algorithm is of order @xmath11 for all @xmath6 . for both measures ,
the upper bounds presented in this paper match lower bounds that were recently obtained by papageorgiou , see @xcite .
hence , the * qs * algorithm enjoys minimal error bounds also in the average - probabilistic setting if we choose @xmath6 divisible by @xmath9 for the first measure and with no restriction on @xmath6 for the second measure .
the quantum setting , and in particular the * qs * algorithm , is relatively new and probably not well known , especially for people interested in continuous complexity .
hence we present all details of this algorithm , emphasizing its quantum parts .
since we wanted also to find sharp error bounds , we needed a very detailed analysis of the outcome probabilities of the * qs * algorithm .
we outline the contents of this paper . in section [ sec : quantsummalgo ] we define the * qs * algorithm .
section [ sec : perfanalys ] deals with the performance analysis of the * qs * algorithm in the worst - probabilistic setting , see section [ sec : wcp ] , and in the average - probabilistic setting , see section [ sec : acp ] .
we consider the most basic form of the summation problem , i.e. , the summation of boolean functions .
let @xmath35 denote the set of boolean functions @xmath36 .
let @xmath37 denote the arithmetic mean of all values of @xmath38 .
clearly , @xmath39 $ ] .
* problem : * for @xmath40 , compute an @xmath41-approximation @xmath42 of the sum @xmath43 such that @xmath44 we are interested in the minimal number of evaluations of the function @xmath38 that are needed to compute @xmath42 satisfying .
it is known that in the worst case setting , we need roughly @xmath45 evaluations of the function @xmath38 . in the randomized setting , we assume that @xmath42 is a random variable and require that holds for the expected value of @xmath46 for any function @xmath38 .
it is known , see e.g. , @xcite , that in the randomized setting we need roughly @xmath47 function evaluations . in the quantum setting , we want to compute a random variable @xmath42 such that holds with a high probability ( greater than @xmath17 ) either for _ all _ boolean functions or on the average with respect to a probability measure defined on the set @xmath35 .
these two error criteria in the quantum setting will be precisely defined in section [ sec : perfanalys ] . in this section
we describe the quantum summation algorithm , which is also called the quantum amplitude estimation algorithm .
this algorithm was discovered by brassard , hyer , mosca and tapp @xcite , and uses grover s iterate operator as its basic component , see @xcite .
we use standard notation of quantum computation , see e.g. , @xcite . for simplicity
we assume that @xmath12 .
let @xmath48 denote the tensor product @xmath49 of @xmath50 copies of @xmath51 , with @xmath51 the 2-dimensional complex vector space .
unit vectors from @xmath51 are called _ one qubit _ quantum states ( or _ qubits _ ) .
let @xmath52 and @xmath53 be an orthonormal basis of @xmath51
. then any qubit @xmath54 can be represented as @xmath55 for @xmath56 , we have @xmath57 , with @xmath58 .
let @xmath59 the set @xmath60 forms an orthonormal basis of @xmath48 and any unit vector @xmath61 can be represented as @xmath62 unit vectors from @xmath48 are called @xmath50 qubit quantum states ( or quantum states or just states , whenever @xmath50 is clear from the context ) .
the only transforms that can be performed on quantum states are defined by certain unitary operators on @xmath48 .
we now define the six unitary operators that are basic components of the summation algorithm .
since unitary operators are linear , it is enough to define them on the basis states @xmath63 . 1
. let @xmath64 denote the _ inversion about zero transform _
@xmath65 where @xmath66 is the kronecker delta .
hence , @xmath67 and @xmath68 for all @xmath69 .
this corresponds to the diagonal matrix with one element equal to @xmath70 , and the rest equal to @xmath71 .
the operator @xmath72 can be also written as the householder operator @xmath73 here , for a state @xmath54 , we let @xmath74 denote the projection onto the space @xmath75 given by @xmath76 where @xmath77 is the inner product in @xmath48 , @xmath78 .
the matrix form of the projector @xmath74 in the basis @xmath79 is @xmath80 .
one can also view the matrix form of the projector @xmath74 as the matrix product of the @xmath81 column vector @xmath54 and the @xmath81 row vector @xmath82 , which is the hermitian conjugate of @xmath54 , @xmath83 . for any @xmath84
we have @xmath85 hence @xmath86 , as claimed .
2 . let @xmath87 denote the _ walsh - hadamard transform _
@xmath88 that is , the walsh - hadamard transform corresponds to the matrix with entries @xmath89 the matrix @xmath90 is symmetric .
furthermore , @xmath91 thus , @xmath92 and @xmath93 is orthogonal .
this means that the operator @xmath94 is symmetric and unitary .
3 . for @xmath95 ,
let @xmath96 denote the _ quantum fourier transform _
@xmath97 hence , @xmath98 corresponds to the unitary block - diagonal matrix @xmath99,\ ] ] where @xmath100 is the matrix of the inverse quantum fourier transform .
for @xmath101 we have @xmath102 the coefficients of @xmath103 in the basis @xmath79 are the inverse quantum fourier transforms of the coefficients of the state @xmath54 .
note that @xmath94 and @xmath104 coincide for the state @xmath52 , i.e. , @xmath105 4 .
let @xmath106 denote the _ quantum query _
operator @xmath107 this again corresponds to the diagonal matrix with elements @xmath108 depending on the values of the boolean function @xmath38 .
this operator is the only one that provides information about the boolean function @xmath38 .
this is analogous to the concept of an _ oracle _ or a _ black - box _ which is used in classical computation and which supplies information about the function @xmath38 through its values .
+ the standard definition of the quantum query @xmath109 is @xmath110 where @xmath111 means addition modulo 2 .
we can simulate @xmath112 by @xmath109 if we use an auxiliary qubit @xmath113 , namely , @xmath114 5 .
let @xmath115 denote the _ grover operator _ @xmath116 this is the basic component of grover s search algorithm , see @xcite .
as we shall see , @xmath117 also plays the major role for the summation algorithm .
the eigenvectors and eigenvalues of @xmath117 will be useful in further considerations .
let @xmath118 and @xmath119 , @xmath120 denote the orthogonal projections of @xmath54 onto the subspaces @xmath121 and @xmath122 , respectively .
that is , @xmath123 then @xmath124 and @xmath125 .
furthermore , @xmath126 , for @xmath127 , so that @xmath128 and @xmath129 , where @xmath130 is the sum we want to approximate .
+ from @xcite , we know that @xmath131 + for the sake of completeness , we provide a short proof of . by the definition of the operator @xmath112 we have @xmath132 and @xmath133 since @xmath134
, we obtain for @xmath127 , @xmath135 from this we calculate for @xmath127 , @xmath136 which is equivalent to .
+ thus , the space @xmath137 is an invariant space of @xmath117 and its eigenvectors and corresponding eigenvalues can be computed by solving the eigenproblem for the @xmath138 matrix @xmath139\,.\ ] ] for @xmath140 , the eigenvectors and the corresponding orthonormalized eigenvalues of @xmath117 are @xmath141 where @xmath142 . moreover , it is easy to check that @xmath143 for @xmath144 , we have @xmath145 and @xmath54 is the eigenvector of @xmath117 with eigenvalues @xmath108 , respectively . for @xmath144 , we define @xmath146 then it is easy to check that ( [ eq : represent ] ) is valid , and @xmath147 is an eigenvalue of @xmath117 for all @xmath148 $ ] .
6 . the next unitary transform , called the _
grover iterate operator _ ,
is defined on the tensor product of @xmath149 and uses @xmath150 qubits .
the first space @xmath151 and @xmath152 qubits will be related to the accuracy of the quantum summation algorithm .
grover iterate _
operator @xmath153 is defined by @xmath154 hence , the power of @xmath117 applied to the second component depends on the first one .
note that @xmath155 may vary from 0 to @xmath156 .
therefore @xmath157 may use the powers of @xmath117 up to the @xmath158st .
we need one more concept of quantum computation , that of _
suppose @xmath159 is a positive integer and consider the space @xmath160 .
given the state @xmath161 we can not , in general , recover all the coefficients @xmath162 .
we can only measure the state @xmath54 with respect to a finite collection of linear operators @xmath163 , where the @xmath164 satisfy the completeness relation @xmath165 after performing the measurement , we obtain the outcome @xmath155 and the state @xmath54 collapses into the state @xmath166 both these events occur with probability @xmath167 .
note that for @xmath168 the outcome @xmath155 can not happen with positive probability .
hence , with probability @xmath71 the outcome @xmath155 corresponds to @xmath169 for @xmath170 .
the most important example of such a collection of operators is @xmath171 .
then , the measurement of the state @xmath54 with respect to this collection of operators gives us the outcome @xmath155 and the collapse into the state @xmath172 with probability @xmath173 , @xmath174 .
another example is a variation of the previous example and will be used in the quantum summation algorithm .
we now let @xmath175 , as for the grover iterate operator , and define @xmath176 by @xmath177 for @xmath178 , with @xmath179 denoting the identity operator on @xmath48 .
that is , @xmath180 for @xmath181 and @xmath182 .
since @xmath183 for all basis states @xmath184 of @xmath151 and @xmath185 of @xmath48 , the completeness relation is satisfied .
consider now the probability of the outcome @xmath155 for a special state @xmath54 of the form @xmath186 with @xmath187 , @xmath188 and @xmath189 for @xmath190 . since @xmath191 is self - adjoint ,
the outcome @xmath155 and the collapse of the state @xmath54 to the state @xmath192 occur with probability @xmath193 .
hence , this collection of operators measures the components of the so - called first register @xmath120 of the quantum state @xmath54 .
following @xcite , we are ready to describe the quantum summation ( * qs * ) algorithm for solving our problem .
the * qs * algorithm depends on a boolean function @xmath38 and on an integer parameter @xmath6 that controls the number of quantum queries of @xmath38 used by the algorithm .
we perform computations in the space @xmath149 , with @xmath194 , so we use @xmath195 qubits .
as we will see later , the accuracy of the algorithm is related to the dimension of the space @xmath151 .
* algorithm qs*(@xmath38 , @xmath6 ) input state : : : @xmath196 with @xmath194 and @xmath197 .
computation : : : 1 .
@xmath198 , 2 .
@xmath199 , 3 .
measurement : : : perform the measurement of the state @xmath201 with respect to the collection @xmath202 . denote the outcome by @xmath155 .
output : : : @xmath203 .
we briefly comment on the * qs * algorithm .
the input state is always the same and does not depend on @xmath38 .
step 1 computes @xmath204 , which is the equally weighted superposition of the basis states .
step 2 computes @xmath205 by using the grover iterate operator . during this step
we use the successive powers of the grover operator @xmath117 , and this is the only step where information about the boolean function @xmath38 is used .
we shall see that the * qs * algorithm uses @xmath206 quantum queries .
step 3 computes @xmath201 by performing the inverse quantum fourier transform on the first @xmath152 qubits , and prepares the system for measurement .
after step 3 , we perform the measurement , obtain the outcome @xmath155 and compute the output @xmath207 on a classical computer .
we stress that the distribution of the outcomes @xmath155 depends on the boolean function @xmath38 , and this is the only dependence of the output @xmath207 on @xmath38 .
in this section we analyze the error of the * qs * algorithm . as we have seen in section [ sec : quantsummalgo ] , the output @xmath207 of the * qs * algorithm is a random value chosen according to a certain distribution dependent on the input function @xmath38 . in this way ,
the * qs * algorithm is a randomized algorithm .
various ways of measuring the performance of randomized algorithms are commonly used in the analysis of algorithms and computational complexity .
they correspond to various error criteria . in this paper
we consider two error criteria : worst - probabilistic and average - probabilistic . in a future paper
we consider other two error criteria : worst - average and average - average , which correspond to the worst or average performance with respect to boolean functions and the average performance with respect to all outcomes .
we start with the error criterion that is used in most papers dealing with quantum computations .
we are interested in the worst case error of the * qs * algorithm that holds with a given probability @xmath15 . here
@xmath208 $ ] and @xmath209 measures the probability of * qs * algorithm s failure and usually @xmath15 is set to be @xmath210 . in our analysis , however , we will allow an arbitrary @xmath211 $ ] .
the choice of the upper bound @xmath212 will be clear from the analysis of the * qs * algorithm .
the * qs * algorithm outputs @xmath207 with probability @xmath213 for @xmath214 , see theorem [ thm : perfanalys ] where the @xmath215 s are given .
its worst - probabilistic error is formally defined as the smallest error bound that holds for all boolean function with probability at least @xmath15 , i.e. , @xmath216 it is easy to see that @xmath217 can be rewritten as follows .
let @xmath218 .
for @xmath219 define the measure of @xmath220 as @xmath221 then @xmath222 the worst - probabilistic error @xmath217 of the * qs * algorithm is defined by the worst performance with respect to boolean functions .
it is also natural to consider the average performance of the * qs * algorithm with respect to boolean functions .
let @xmath223 be a probability measure on the set @xmath35 .
that is , any boolean function @xmath40 occurs with probability @xmath224 .
obviously , @xmath225 and @xmath226 .
the average - probabilistic error is defined by replacing the first max in by the expectation , i.e. , @xmath227 hence , we are interested in the average error that holds with a certain fixed probability .
we begin by citing a theorem from @xcite for which we will propose a number of improvements .
[ thm : brass&co ] @xcite for any boolean function @xmath40 , the * qs * algorithm uses exactly @xmath1 quantum queries and outputs @xmath228 that approximates @xmath130 such that @xmath229 with probability at least @xmath230 . hence , @xmath231 and , therefore ,
@xmath232 using proof ideas of theorem [ thm : brass&co ] from @xcite we present the following theorem and the subsequent corollaries .
[ thm : perfanalys ] for any boolean function @xmath233 , denote @xmath234.\ ] ] 1 .
the * qs * algorithm uses exactly @xmath1 quantum queries , and @xmath235 qubits .
2 . for @xmath236 , the outcome @xmath155 of the * qs * algorithm occurs with probability @xmath237 ( if @xmath238 we need to apply the limiting value of the formula above . ) for @xmath239 , the outcome @xmath155 occurs with probability @xmath240 .
3 . if @xmath241 is an integer then the * qs * algorithm outputs the exact value of @xmath43 with probability @xmath71 . this holds iff @xmath242 for some integer @xmath243 $ ] .
in particular , this holds for @xmath244 , for @xmath245 and even @xmath6 , and for @xmath246 and @xmath6 divisible by @xmath9 .
4 . let @xmath247 and @xmath248 . if @xmath241 is not an integer then the * qs * algorithm outputs the same value @xmath249 for the outcomes @xmath250 and @xmath251 such that @xmath252 with probability @xmath253 and outputs the same value @xmath254 for the outcomes @xmath255 and @xmath256such that @xmath257 with probability @xmath258 as before , let @xmath259 and @xmath260 for arbitrary @xmath261 .
note that @xmath262 the steps 13 of the * qs * algorithm are equivalent to the application of the operator @xmath263 to the state @xmath264 .
then @xmath265 can be written as @xmath266 , and @xmath267 is given by . hence @xmath268 applying @xmath157 in step 2 and remembering that @xmath269
, we obtain @xmath270 since @xmath271 , the largest power of @xmath117 is @xmath1 .
hence , we use exactly @xmath1 quantum queries to compute @xmath205 .
the remaining steps of the * qs * algorithm do not use quantum queries .
this means that the total number of quantum queries used by the algorithm is @xmath1 , and obviously we are using @xmath195 qubits .
this proves the first part of theorem [ thm : perfanalys ] .
step 3 yields the state @xmath272 we are ready to analyze the probability of the outcome @xmath155 of the * qs * algorithm .
observe that @xmath273 and therefore @xmath274 observe that for @xmath140 , we have @xmath275 , whereas for @xmath144 , we have @xmath276 and @xmath277 . for @xmath278
we have @xmath279 if @xmath280 is an integer then the last sum is clearly @xmath6 , and the whole expression is @xmath71 .
if @xmath280 is not an integer then @xmath281 which holds for all @xmath282 if we take @xmath283 as @xmath71
. therefore @xmath284 hence @xmath285 which holds for all @xmath282 if we take @xmath283 as @xmath71 . applying this
we conclude that @xmath286 the outcome @xmath155 occurs after the measurement and the state @xmath201 collapses to the state @xmath287 , where @xmath288 . for @xmath214
, we have @xmath289 whereas @xmath290 for @xmath291 .
since @xmath292 and @xmath293 are orthogonal we have @xmath294 hence , the outcome @xmath155 , @xmath236 , occurs with probability @xmath295 indeed , for @xmath140 , we have @xmath296 and ( [ 158 ] ) follows from the form of @xmath297 . for @xmath144 , we have @xmath298 and @xmath277 .
since the two terms in ( [ 158 ] ) are now the same , the formula for @xmath299 again yields ( [ 158 ] ) .
since @xmath300 , the last formula is equivalent to ( [ eq : pj ] ) . obviously for @xmath301 , the probability of the outcome @xmath155 is zero .
this proves the second part of theorem [ thm : perfanalys ] .
assume now that @xmath302 . if @xmath303 or @xmath304 ( if @xmath6 is even ) then the probability @xmath305 of the outcome @xmath306 is @xmath71 .
for @xmath303 we have @xmath307 and the output is @xmath308 .
for @xmath304 we have @xmath309 and the output is @xmath310 .
hence , in both cases the * qs * algorithm outputs the exact value with probability @xmath71 .
if @xmath311 and @xmath312 then the probability of the distinct outcomes @xmath306 and @xmath313 is @xmath17 .
these two values of the outcomes yield the same output @xmath314 hence , the * qs * algorithm outputs the exact value with probability @xmath71 .
this proves the third part of theorem [ thm : perfanalys ] .
we now turn to the case when @xmath315 .
it is easy to check that the third part of theorem [ thm : perfanalys ] holds for @xmath316 .
assume then that @xmath317 which implies that @xmath318 .
since @xmath306 is not an integer , we have @xmath319 , @xmath320 and @xmath321 .
this means that both @xmath250 and @xmath251 may be the outcomes of the * qs * algorithm .
obviously , these two outcomes are different iff @xmath322 .
similarly , both @xmath255 and @xmath323 may be also the outcomes .
they are different iff @xmath324 .
we show that the outputs for the outcomes @xmath250 and @xmath255 satisfy and with probabilities and , respectively .
we focus on the output for the outcome @xmath250 and its probability .
the proof for the outcome @xmath255 is similar .
we estimate the error of the * qs * algorithm for the output @xmath325 .
recall that @xmath326 .
we have @xmath327 since @xmath328 and @xmath329 , this proves the estimate of the error of the * qs * algorithm in the fourth part of theorem [ thm : perfanalys ] .
we find the probability of the output @xmath228 .
since @xmath330 is injective for @xmath331 $ ] , the output @xmath228 occurs only for the outcomes @xmath250 and @xmath251 .
if @xmath332 then these two outcomes are the same and @xmath228 occurs with probability @xmath333 . due to ( [ 158 ] ) @xmath334 which agrees with the claim in theorem [ thm : perfanalys ] .
if @xmath335 then @xmath336 and @xmath228 occurs for exactly two distinct outcomes .
the probability of @xmath228 is now equal to the sum of the probabilities @xmath337 with @xmath338 s given by ( [ 158 ] ) . since both terms are equal , the probability of @xmath228 is @xmath339 which also agrees with the claim in theorem [ thm : perfanalys ] .
since @xmath340 we have @xmath341 we finish proving using the standard expansion of the sine .
this completes the proof .
based on theorem [ thm : perfanalys ] we present simplified estimates of the error of the * qs * algorithm and of the corresponding probability .
[ maxcor ] the * qs * algorithm outputs @xmath228 such that @xmath342 with probability at least @xmath343 it is enough to prove corollary [ maxcor ] if @xmath306 is not an integer .
the estimate of the error of the * qs * algorithm by the maximum of the estimates and holds with probability that is the sum of the probabilities and .
moreover , @xmath344 .
it now suffices to observe that the function @xmath345 is a lower bound of the left hand side of with @xmath346 , and attains the minimum @xmath3 on the interval @xmath347 $ ] for @xmath348 , see also @xcite .
corollary [ maxcor ] guarantees high probability of the estimate . unfortunately
this estimate does not preserve the continuity of the estimates and with respect to @xmath349 and @xmath350 .
the continuity of the estimates will be present in the next corollary at the expense of the probability of the outcome .
this corollary will also play an essential role in the study of the average - probabilistic error of the * qs * algorithm .
[ mincor ] the * qs * algorithm outputs @xmath228 such that @xmath351 with probability at least @xmath352 we may again assume that @xmath306 is not an integer .
let us define @xmath353.\ ] ] then @xmath354 is the probability of and @xmath355 is the probability of . for @xmath356
$ ] , note that @xmath357 is decreasing , and @xmath358 is increasing
. therefore @xmath359.\ ] ] let @xmath360 . then @xmath361 . in this case
is equivalent to and holds with probability at least @xmath362 , which corresponds to .
analogously , if @xmath363 then @xmath364 . in this case
is equivalent to and holds with probability at least @xmath365 , which also corresponds to .
finally , note that @xmath366 is minimal for @xmath367 and is equal to @xmath368 unfortunately for @xmath369 close to @xmath17 the probability of the estimate is too small .
however , in this case we may use corollary [ maxcor ] , which yields the estimate with high probability .
we now turn to global error estimates , that is , estimates independent of @xmath370 .
theorem [ thm : brass&co ] of @xcite states , in particular , that @xmath371 with probability at least @xmath3 .
we now improve this estimate by combining the estimates and .
[ improvedcor ] the * qs * algorithm outputs @xmath228 such that @xmath372 with probability at least @xmath3 .
that is , @xmath373 let us define @xmath374 clearly , @xmath375 is a lower bound of the @xmath376 and therefore @xmath377 is a lower bound of the probability of the output satisfying ( [ eq : mincorerr ] ) .
we consider two cases .
assume first that @xmath378 \cup [ \tfrac{3}{4},1]$ ] .
it is easy to see that then @xmath379 and the estimate yields @xmath380 with probability at least @xmath3 .
assume now that @xmath381 .
then we can use the estimate , which holds unconditionally with probability at least @xmath3 . in this case , we have @xmath382 these two estimates combined together yield .
the obvious consequence of corollary [ improvedcor ] is that for @xmath6 large enough we can compute the value of @xmath370 exactly by rounding the output .
assume that @xmath383 then the rounding of the * qs * algorithm output to the nearest number of the form @xmath384 yields the exact value of the sum @xmath370 with probability at least @xmath3 .
the proof of corollary [ improvedcor ] may suggest that the constant @xmath210 in can be decreased .
furthermore one may want to decrease the constant @xmath210 at the expense of decreasing the probability @xmath3 .
these points are addressed in the next corollary .
we shall see that the constant @xmath210 may be lowered only by decreasing the probability .
[ globalcor ] define @xmath385 and @xmath386.\ ] ] then @xmath387 , \\
m/\pi & \text{for } \quad p \in ( 8/ \pi^2 , 1 ] .
\\ \end{cases}\ ] ] moreover , @xmath388 $ ] and @xmath389.\ ] ] for @xmath390 , corollary [ globalcor ] is a consequence of corollary [ mincor ] . for @xmath391
$ ] , corollary [ globalcor ] trivially holds since @xmath392 . for the remaining @xmath15
s we use a proof technique similar to that of corollary [ improvedcor ] . let @xmath393 $ ] .
it is easy to check that @xmath394 is decreasing and , therefore , @xmath395 is well defined and @xmath396 $ ] .
we have to show that the estimate @xmath397 holds with probability at least @xmath15 .
we consider two cases .
assume first that @xmath398 \cup [ 1-v^{-1}(p),1]$ ] .
observe that the function @xmath399 defined in can be rewritten as @xmath400 it is easy to see that in this case , @xmath401 and the estimate yields @xmath402 with probability at least @xmath15 .
assume now that @xmath403 .
then we can use the estimate , which holds unconditionally with probability at least @xmath404 . in this case
, we have @xmath405 this proves .
we found the estimate by numerical computations .
for @xmath406 $ ] ] from figure [ fig : cp ] we see that the estimate is almost linear on the interval @xmath407 $ ] , which explains why the right hand side of the estimate is small . on @xmath408 $ ] .
the two horizontal lines show @xmath409 and @xmath3 levels .
the part of the graph between the arrows shows that @xmath394 is almost linear . ]
we now find a sharp bound on the worst - probabilistic error of the * qs * algorithm .
we show that for large @xmath6 and @xmath24 the bound obtained in corollary [ globalcor ] is optimal for @xmath25 $ ] .
[ thm : asympp ] for large @xmath6 and @xmath24 , the worst - probabilistic error of the * qs * algorithm is given by @xmath410.\ ] ] here , @xmath394 is as in corollary [ globalcor ] , and @xmath411 by ( [ eq : linearestim ] ) . from corollary [ globalcor ]
, it is enough to show a lower bound on the error .
define @xmath412 for large @xmath6 , we have @xmath413 there exist two boolean functions @xmath414 and @xmath415 with sums @xmath416 and @xmath417 such that @xmath418 since @xmath419 and the derivative of @xmath306 for @xmath420 is @xmath421 , we have @xmath422 obviously , @xmath423 for some integers @xmath424 with @xmath425 .
consider @xmath426 for @xmath427 .
then @xmath426 varies from @xmath428 for @xmath429 to @xmath430 for @xmath431 . since @xmath432 , for a positive and small @xmath433 ( we finally let @xmath433 go to zero ) , we can choose @xmath434 such that for @xmath435 we have @xmath436 for large @xmath24 , we then have @xmath437 let @xmath438 denote the output for the outcome @xmath439 , and @xmath440 for @xmath441 .
due to ( [ eq : perfanalyserr1 ] ) and ( [ eq : perfanalyserr2 ] ) of theorem [ thm : perfanalys ] we have @xmath442 let us write @xmath443 for @xmath444 .
the probability of @xmath440 is given by ( [ eq : perfanalyspr2 ] ) and is now equal to @xmath445 since @xmath446 , the probability of @xmath447 is @xmath448 since @xmath449 for @xmath450 $ ] , we see that the probability of @xmath447 is slightly less than @xmath15 for small @xmath433 .
we are ready to find a lower bound on the worst - probabilistic error @xmath451 of the * qs * algorithm .
take the function @xmath38 that corresponds to @xmath452 .
we claimed that the error is minimized if @xmath453 .
indeed , @xmath441 must belong to @xmath220 since otherwise @xmath454 for @xmath21 .
the probability of @xmath441 is slightly less than @xmath15 , and so the set @xmath220 must also contain some other outcome @xmath155 .
if @xmath455 then the error bound is roughly @xmath456 , and the sum of the probabilities of the outputs for the outcome @xmath441 and @xmath439 is always at least @xmath457 . on the other hand , if @xmath439 does not belong to the set @xmath220 then any other outcome @xmath155 yields the output @xmath458 .
since @xmath459 , the distribution of the outcomes around @xmath17 is a mesh with step size roughly @xmath460 .
hence , if @xmath461 , the error is at least roughly @xmath462 .
thus the choice @xmath455 minimizes the error and for @xmath433 tending to zero , the error is roughly @xmath463 .
this completes the proof . from these results , it is obvious how to guarantee that the error of the * qs * algorithm is at most @xmath41 with probability at least @xmath15 .
since @xmath464 holds with probability @xmath15 , it is enough to take @xmath465 . due to theorem [ thm : asympp ] this
bound is sharp for small @xmath41 and large @xmath466 .
we have [ epsiloncor ] for @xmath25 $ ] , the algorithm * qs*(@xmath38 , @xmath467 ) computes @xmath228 with the error @xmath41 and probability at least @xmath15 with @xmath468 quantum queries . for small @xmath41 and large @xmath466 ,
the estimate of the number of quantum queries is sharp . in this section
we study the average performance of the * qs * algorithm with respect to some measure on the set @xmath35 of all boolean functions defined on the set @xmath469 .
we consider two such measures .
the first measure @xmath470 is uniformly distributed on the set @xmath35 , i.e. , @xmath471 the second measure @xmath472 is uniformly distributed on the set of results , i.e. , @xmath473 for the average - probabilistic error we want to estimate @xmath474 for the measures @xmath475 , the mean of the random variable @xmath43 is clearly @xmath17 . however , their first ( central ) moments are very different .
as we shall see , the moment for the measure @xmath470 is small since it is of order @xmath32 whereas the moment for measure @xmath472 is roughly @xmath33 .
since the first moments are the same as the error of the constant algorithm @xmath476 , we can achieve small error of order @xmath32 for the measure @xmath470 without any quantum queries , while this property is not true for the measure @xmath472 .
we now consider the measure @xmath470 .
it is interesting to ask if the * qs * algorithm has the same property as the constant algorithm .
we shall prove that this is indeed the case iff @xmath6 is divisible by 4 .
we compute the first moment or the error of the constant algorithm , which is @xmath477 we do it only for odd @xmath0 since the case of even @xmath0 is analogous .
we have @xmath478 thus @xmath479 by stirling s formula @xmath480 $ } , \ ] ] we estimate the both binomial quantities in ( [ eq : zero_err ] ) by @xmath481 proving that @xmath482 we are ready to analyze the average - probabilistic error of the * qs * algorithm .
[ thm : wa4 ] assume that @xmath6 is divisible by @xmath9 and let @xmath483 $ ] .
then the average - probabilistic error of the * qs * algorithms with respect to the measure @xmath470 satisfies @xmath484 the estimate @xmath485 is obvious and , applying corollary [ improvedcor ] , we get @xmath486 as before denote @xmath487
. let @xmath488 .
we are interested in the behavior of @xmath489 for @xmath490 .
clearly @xmath491 .
let @xmath492 , by taylor s theorem , we have @xmath493 and @xmath494 .
assume additionally that @xmath495 which is equivalent to assuming that @xmath496 since @xmath6 is divisible by @xmath9 then @xmath497 for @xmath498 , and @xmath499 for @xmath500 .
this yields @xmath501 observe that @xmath502 \cup [ \tfrac{3}{4},1].\ ] ] indeed , for @xmath503 we have @xmath504,\ ] ] and for @xmath505 we have @xmath506=[\tfrac{3}{4},1],\ ] ] as claimed .
let @xmath507 . by the proof of corollary [ improvedcor ] ,
the error of the * qs * algorithm satisfies @xmath508 and by ( [ eq : min_estim ] ) , we have @xmath509 we split the sum that defines @xmath510 into two sums .
the first sum is for @xmath219 for which @xmath511 with @xmath512 and the second sum is for @xmath38 for which @xmath513 with @xmath514 .
for the first sum we estimate the error of the * qs * algorithm by @xmath515 and for the second sum , by the worst - case error @xmath516 .
hence we have @xmath517 since @xmath518 for some integer @xmath519 $ ] , @xmath520 since @xmath521 , the first sum can be estimated as @xmath522 the second sum can be estimated by @xmath523 adding the estimates of these two sums we obtain @xmath524 the last sum is given by ( [ eq : zero_err ] ) and estimated by ( [ needed ] ) .
hence @xmath525 which completes the proof . in the next theorem
we consider the case when @xmath6 is not divisible by 4 .
assume that @xmath526 is not divisible by @xmath9 , and let @xmath527 $ ] .
then the average - probabilistic error of the * qs * algorithm with respect to the measure @xmath470 satisfies @xmath528 let @xmath529 for @xmath530 .
let , as before , @xmath531 .
as in the proof of theorem [ thm : wa4 ] , for @xmath532 we have @xmath533 and @xmath494 . assume additionally that @xmath534 which is equivalent to assuming that @xmath535 thus for @xmath536 satisfying ( [ eq : xestim ] ) we have @xmath537,\ ] ] and @xmath538 and @xmath539 . from the proof of corollary [ maxcor ] we have @xmath540 .
since @xmath541 then either @xmath542 or @xmath543 .
we then estimate @xmath544 we now estimate the error of the * qs * algorithm for @xmath545 such that @xmath546 for @xmath536 satisfying ( [ eq : xestim ] ) and the outcome @xmath547 or @xmath548 . denote the outcome by @xmath549 for @xmath550 . by taylor
s theorem we have @xmath551$}.\end{gathered}\ ] ] since @xmath552 for @xmath553 $ ] , we have @xmath554 .
consider the error for the outcome @xmath549 and @xmath536 satisfying ( [ eq : xestim ] ) .
then @xmath555 and the error can be estimated by @xmath556 clearly , @xmath557 and @xmath558 $ ]
. then @xmath559 .
hence , for @xmath38 such that @xmath561 with @xmath536 satisfying ( [ eq : xestim ] ) we have @xmath562 we are now ready to estimate @xmath563 .
first , by ( [ eq : eap_estim ] ) , we have @xmath564 this , ( [ eq:123 ] ) and the bernstein inequality , @xmath565 yields @xmath566 which completes the proof .
obviously , in the average - probabilistic setting , we should use the * qs * algorithm with @xmath6 divisible by @xmath9 .
then theorem [ thm : wa4 ] states that the error is of order @xmath34 .
recently , papageorgiou @xcite proved that for any quantum algorithm that uses @xmath6 quantum queries the error is bounded from below by of @xmath567 with probability @xmath25 $ ] . here
, @xmath568 is a positive number independent of @xmath6 and @xmath0 .
hence , the * qs * algorithm enjoys an optimality property also in the average - probabilistic setting for the measure @xmath470 as long as we use it with @xmath6 divisible by @xmath9 .
we now turn to the measure @xmath472 . clearly , the average - probabilistic error of the * qs * algorithm is bounded by its worst - probabilistic error , which is of order @xmath11 with probability @xmath483 $ ] .
it turns out , again due to a recent result of papageorgiou @xcite that this bound is the best possible , since any quantum algorithm that uses @xmath6 quantum queries must have an error proportional at least to @xmath11 .
hence , the factor @xmath32 that is present for the measure @xmath470 does not appear for the measure @xmath472 , and the behavior of the * qs * algorithm is roughly the same in the worst- and average - probabilistic settings .
we are grateful for many valuable comments on this paper from j. creutzig , a. papageorgiou , j. f. traub , and a.g .
werschulz .
l. grover , a fast quantum mechanical algorithm for database search , proceedings of the 28th annual acm symposium on the theory of computing ( stoc ) , 212 - 219 , 1996 , http://arxiv.org/quant-ph/9605043 , 1996 .
a. nayak and f. wu , the quantum query complexity of approximating the median and related statistics , proceedings of the 31th annual acm symposium on the theory of computing ( stoc ) , 384 - 393 , 1999 , http://arxiv.org/quant-ph/9804066 , 1998 . | we study the quantum summation ( * qs * ) algorithm of brassard , hyer , mosca and tapp , see @xcite , that approximates the arithmetic mean of a boolean function defined on @xmath0 elements .
we improve error bounds presented in @xcite in the worst - probabilistic setting , and present new error bounds in the average - probabilistic setting . in particular , in the worst - probabilistic setting , we prove that the error of the * qs * algorithm using @xmath1 quantum queries is @xmath2 with probability @xmath3 , which improves the error bound @xmath4 of @xcite .
we also present error bounds with probabilities @xmath5 $ ] , and show that they are sharp for large @xmath6 and @xmath7 . in the average - probabilistic setting , we prove that the * qs * algorithm has error of order @xmath8 iff @xmath6 is divisible by @xmath9 .
this bound is optimal , as recently shown in @xcite . for @xmath6
not divisible by @xmath9 , the * qs * algorithm is far from being optimal if @xmath10 since its error is proportional to @xmath11 . |
during the last years a spectacular development in the storage and manipulation of cold atoms in optical lattices @xcite has taken place .
greiner et al .
@xcite , to name one important example , succeeded in experimentally driving a quantum phase transition between a superfluid and a mott - insulating phase in bosonic systems . this experimental progress has revived the interest in the bose - hubbard model [ eq .
( [ bh ] ) ] as a generic hamiltonian for strongly correlated bosons , by which the quantum phase transition can be described @xcite .
the bose - hubbard hamiltonian has been used previously in condensed matter physics to study the adsorption of noble gases in nanotubes @xcite , or cooper pairs in superconducting films with strong charging effects @xcite . in this context
a lot of work has already been done to characterize the quantum phase transition , the statistics , and the low - energy excitations of the bose - hubbbard model @xcite .
however , new interesting questions arise now due to the good tunability of the experiments with optical lattices . in particular
, it becomes possible to study time - dependent processes such as driven quantum phase transitions @xcite .
a theoretical understanding of such phenomena is challenging , since the characteristics of the superfluid phase where the atoms tend to delocalize throughout the lattice and large fluctuations in the local density exist , and the mott - insulating phase where the number fluctuations decrease , and a gap in the excitation spectrum opens must be covered at the same time .
both regions are separated by a non - analyticity of the spectrum , which implies that a perturbative study @xcite works best in strong coupling limit , while a hartree - fock - bogoliubov mean field works best in the superfluid regime .
in addition it is possible to develop a mean field theory @xcite based on a gutzwiller ansatz @xcite : this reproduces the mean field theory in the superfluid limit , as well as the limit of infinite interaction , which raises the hope that the theory also interpolates properly between these limits . in this paper
we develop a variational description of the ground state of an ensemble of cold atoms in an one - dimensional optical lattice .
our trial wavefunction treats the connections between neighboring sites as entities which decouple in the limit of infinitely large lattices .
we apply this technique first to the quantum rotor model which describes the lattice for large and commensurate occupation per site @xcite , and has also been used to describe an array of josephson junctions @xcite , and next to the bose - hubbard hamiltonian .
the accuracy of the variational theory in both models is confirmed by several comparisons . in the quantum rotor case ,
we use a spin wave approximation in the limit of weak interaction , and a first order perturbation theory in the limit of weak tunneling . for the bose - hubbard model
we compare against results obtained applying the quasi - exact , numerical dmrg method to one - dimensional lattices with up to 128 sites , and also with calculations based on the gutzwiller ansatz .
our conclusion is that the variational picture of self - regulated connections between sites provides a rather cheap and simple , but very good description of the local properties of the system in the superfluid and insulator regimes , and a fairly good interpolation across the quantum phase transition .
it can not describe , however , the algebraic decay with distance of the off - diagonal elements of the one - particle density matrix .
the simplicity of the method suggests a possible generalization to higher dimensionalities and other physical models .
the outline of the paper is as follows : in sec .
[ sec - rotor ] we introduce the quantum rotor model as a possible limit of the bose - hubbard hamiltonian .
next , information about the ground state of the quantum rotor model is obtained variationally as the solution of a mathieu equation .
we can estimate energies , correlation functions and length , and the variance of the density as a function of the only free parameter .
a comparison with perturbative estimates demonstrates the accuracy of the method when computing local properties . since the quantum rotor model is only an approximate description of the optical lattice , in sec .
[ sec - coherent ] we develop a similar variational theory for the bose - hubbard hamiltonian .
after bringing the bose - hubbard hamiltonian to an appropriate form , we can estimate the local properties of its ground state .
the variational solutions are compared in sec .
[ sec - dmrg ] with the results of dmrg studies of the bose - hubbard model .
we confirm that the variational method describes very well the local properties of both the mott - insulator and the superfluid regime , and provides a fairly good interpolation across the phase transition . finally , in sec .
[ sec - end ] we summarize our results and comment on possible extensions .
in this section we show the equivalence of the bose - hubbard model @xmath0 - \frac{u}{2 } m \bar{n}(\bar{n}-1),\label{bh}\ ] ] and the quantum rotor model for large and integer occupation @xmath1 @xcite . in our notation , @xmath2 is the number of lattice sites and @xmath3 the number of atoms .
both the bose - hubbard model and the quantum rotor model show a phase transition due to the interplay between the kinetic term proportional to @xmath4 and the interaction term proportional to @xmath5 . for convenience we have subtracted the ground state energy in the perfect insulator limit @xmath6 .
if we expand a configuration of the lattice using fock states @xmath7 and the number of particles per lattice site is large , @xmath8 , we may approximate the hopping terms as follows @xmath9 here , @xmath10 are ladder operators and @xmath11 projects on states with non - negative occupation numbers , @xmath12 .
to lowest order the error @xmath13 is @xmath14 and its norm is bound by @xmath15 where @xmath16 is the variance in the number of particles per lattice site . for the approximation ( [ aprox - kin ] ) to be valid , the uncertainty in the number of atoms must be small compared to the mean value , @xmath17 , and the interaction energy must exceed the neglected terms , @xmath18 .
following the previous procedure the bose - hubbard model becomes @xmath19 \label{h - qr}\ ] ] here @xmath20 is approximately the density , @xmath21 is essentially the number operator and we have used that @xmath22 when we work with states that have a fixed , commensurate number of particles . in the following
we will define the energy per lattice site as @xmath23 since the physically interesting states will be concentrated around large occupations , @xmath24 , the usual step now is to drop the projector , @xmath11 , and move to the basis of phase states , defined by @xmath25^{\otimes m}. \label{def-1}\ ] ] in doing so , we obtain the identification @xmath26 and @xmath27 , which produces the usual representation of the quantum rotor model @xmath28 , \label{h - qr-2}\ ] ] with the associated state writen as @xmath29 a similar derivation is possible using path integrals @xcite . in this section
we estimate the properties of the ground state of @xmath30 variationally . due to the previous splitting ( [ phase ] ) ,
any wavefunction @xmath31 can only depend on the phase difference between neighboring wells , @xmath32 .
these new quantum variables describe the connections between neigboring sites . in the limit of large lattices it seems reasonable to assume that these connections become independent from each other adopting the product state @xmath33
this representation becomes exact in the mott - insulating limit , @xmath34 , where @xmath35 , and in the superfluid limit , @xmath36 , where @xmath37 , as can be verified by direct substitution in eq .
( [ phase ] ) . even though the phase representation is the best one to find a trial wavefunction , it is not the optimal one for performing computations .
it is instead more convenient to work with the variables which are conjugate to the phase differences @xmath32 .
these are the new quantum numbers , @xmath38 , given by @xmath39 in terms of these numbers , the ansatz ( [ ansatz - phase ] ) reveals itself as a simple product wavefunction @xmath40 with coefficients given by the fourier transform @xmath41 as sketched in fig .
[ fig - m - numbers ] , the @xmath38 play the roles of chemical potentials which are established between different wells : the difference between the potentials on the extremes of a site gives the fluctuations over the mean and commesurate occupation @xmath1 . in this picture @xmath42 , \label{h - qr-3}\ ] ] where @xmath43 are new infinite - dimensional ladder operators and @xmath44 . by minimizing the energy associated with @xmath30 over all states within a given ansatz
we can both obtain an upper bound to the energy of the ground state and approximate its wave function .
a simple computation with our product ansatz leads to the result [ compare ( [ eq : mu ] ) ] @xmath45 \simeq
-4j\rho \mathrm{re}{\langle{\sigma^+}\rangle } + u{\langle{(\sigma^z)^2}\rangle } - u { \langle{\sigma^z}\rangle}^2,\ ] ] where the expected values are computed over a single connection , @xmath46 , and the wavefunctions are assumed to be normalized , @xmath47 . since the stationary states have a well defined parity , @xmath48 , the optimal variational state must satisfy the linear equation @xmath49 which is nothing but the fourier transform of a mathieu equation @xmath50
h(\xi ) = \varepsilon_{{est } } h(\xi),\label{mathieu}\ ] ] the estimated ground state energy per site is given by the lowest eigenvalue of either equation .
using the product ansatz and the same approximations required to derive @xmath30 , we can also compute other properties of the ground state .
for instance , the variance of the number of atoms per lattice site @xmath51 and the correlation functions @xmath52 which decay exponentially with the distance .
this implies that the ansatz ( [ ansatz ] ) only describes properly the decay of the correlations in the mott - insulating regime , since the correlations in the superfluid regime follow a power law decay .
however , as we will see below , local magnitudes ( @xmath53 ) are properly estimated even if long range ones are not .
we have solved eq .
( [ mathieu - discrete ] ) numerically in a truncated space .
the results are summarized in fig .
[ fig - qr ] , where we also plot reference estimates arising from two other analytical methods . in the limit @xmath54
we compare with a first order perturbative calculation around the solution @xmath55 , which is possible thanks to the energy gap of order @xmath56 in the excitation spectrum . in the limit @xmath57
we rather use a spin wave or harmonic approximation in which the cosine - term of the hamiltonian @xmath30 is expanded up to second order in the phase difference between neighboring sites [ see sec .
[ appendix - c ] ] .
this approximation is valid in the superfluid regime , where the phase does not vary much between neighboring wells . from the graphical comparison
we see that , as expected , the variational wavefunction provides a fairly accurate description of the ground state of the quantum rotor model in both the superfluid and insulating limits . as a side note
, we must remark that this ground state has a divergent fluctuation of the number of particles per site as @xmath58 .
this disagrees from the expected behavior of the ground state of the original bose - hubard hamiltonian , and it reminds us that @xmath30 can only model the atomic ensemble when the variance , @xmath59 , is small compared to the mean occupation number , @xmath1 . in the limit @xmath60 it is possible to estimate the ground state of the rotor model ( [ h - qr ] ) analytically .
since we are deep in the superfluid regime , the wavefunction will be concentrated around the line @xmath61 , and we can approximate @xmath62.\ ] ] if we remove the periodic boundary conditions on @xmath63 and change variables , the preceding hamiltonian may be diagonalized , @xmath64 , with frequencies given by @xmath65 , where @xmath66 is an integer in the range @xmath67 which labels the different values of the momentum in the lattice .
the ground state energy [ fig .
[ fig - gutz](a ) ] may be estimated as the zero - point energy of the harmonic oscillator .
for large @xmath2 , the sum over @xmath66 may be replaced with an integral , giving @xmath68 the variance of the number of particles is related to the expectation value of the momentum using the same procedure as above @xmath69
in this section we apply to the bose - hubbard hamiltonian the techniques that were developed in sec .
[ sec - rotor ]
. we will do it in three steps : first we will develop a phase representation which is valid for all occupation numbers . next we will prove that this representation is equivalent to a similarity transformation of the hamiltonian which brings it to a form similar to ( [ h - qr ] ) , at the price of losing hermiticity .
finally we will show how to implement the ansatz of independent connections ( [ ansatz ] ) to produce estimates for the usual set of observables ( @xmath70 ) , which are to be validated with dmrg calculations . the phase coherent states @xmath71 are defined by @xmath72 unlike the phase states defined in sec .
[ sec - rotor ] , they are not orthogonal to each other , @xmath73 $ ] , but they form a complete basis , so that an expansion like ( [ phase ] ) is still possible .
a nice property of the coherent states is that we can rewrite the operators @xmath74 , @xmath75 , @xmath76 , etc , in terms of the phases very easily .
for instance , @xmath77 , @xmath78 , and @xmath79 . using this representation ,
we obtain an effective hamiltonian for the wavefunction @xmath31 , i.e. @xmath80 { |{\vec \phi}\rangle}$ ] , which is of the form @xmath81 + \frac{u}{2}\sum_j \left(-\frac{\partial^2}{\partial\phi_j^2}\right ) . \label{h - phase-2}\ ] ]
here @xmath82 stands for the tranpose of @xmath83 .
this operator was already used in ref .
@xcite to study the bose - hubbard model with only two sites . on the one hand
, it is a non - hermitian operator is maintained due to an implicit projection that takes place when we reconstruct the state @xmath84 from @xmath85 ( see ref .
@xcite ) . ] and we can not do a simple variational study . on the other hand
the hamiltonian still depends on the phase differences , and it is reasonable to look for approximate eigenstates which have the form ( [ ansatz - phase ] ) .
this will be done in the following section . however , since working with phase variables is inconvenient , we will develop a representation similar to that of eq .
( [ h - qr-3 ] ) in the following . in this section
we will find the best variational function which has the product form of eq .
( [ ansatz - phase ] ) .
however , as it happened in sec . [ sec - rotor ] , instead of working with phase variables it is more convenient develop a representation of the bose - hubbard hamiltonian in terms of connections .
this is once more a two - steps process .
first we use a similarity transformation suggested by the definition of the coherent states @xmath86 since @xmath87 and @xmath88 , we find @xmath89 the hamiltonians ( [ h - phase-2 ] ) and ( [ h - phase ] ) are equivalent : while one is defined in terms of phase variables , the other one is defined with occupations numbers , and both are related by a fourier transform . the second and final step is to rewrite everything in terms of connections , using the quantum numbers from eq .
( [ quantum - numbers ] ) , and the relations @xmath90 , @xmath91 .
the result is a decomposition of the hamiltonian @xmath92,\nonumber\\ h_2 & = & \sum_j\left[j(\sigma^z_{j-1}\sigma^+_j-\sigma^z_{j+1}\sigma^-_j)+ u \sigma^z_j\sigma^z_{j+1}\right],\nonumber\end{aligned}\ ] ] into terms which are local , @xmath93 , and terms which involve pairs of connections , @xmath94 . for the quantum rotor model we proved that the optimal product wavefunction was an eigenstate of a hamiltonian which did not couple connections , like @xmath93 .
the difference now is that , since the operator @xmath83 is not hermitian , we can not establish a variational principle and that proof is no longer valid .
nevertheless , we will again propose a variational ansatz which is an eigenstate of the local operator @xmath95 .
using the following equality @xmath96 and the product ansatz @xmath97 , we arrive to an upper bound for the lowest eigenvalue of the bose - hubbard hamiltonian , expressed in terms of the non - hermitian one @xmath98 the way to use this variational principle is as follows .
first , for a given @xmath4 and @xmath5 we compute the lowest eigenstate of @xmath93 and this way obtain @xmath99 . after the equivalence ( [ equivalence ] ) , finding the ground state of the local hamiltonian @xmath93 becomes equivalent to solving a modified mathieu equation @xmath100h = \varepsilon_{{est}}h,\ ] ] which describes exactly the static properties of a pair of sites with open boundary conditions @xcite .
once we have @xmath101 we must still compute the correction @xmath102 using a rather straightforward expansion which is shown in sec .
[ sec - bounds ] .
surprisingly , @xmath102 happens to be negative , so that it is actually an improvement over the simple estimate given by @xmath101 [ see fig .
[ fig - correction ] ] . from the optimal variational state ,
@xmath103 , and the estimate for the energy , @xmath104 , we may compute other observables .
for the density fluctuations and nearest neighbor correlations we use the virial theorem @xmath105 whereas for other properties one has to evaluate numerically the matrix products shown in sec .
[ sec - bounds ] .
this allows us to prove that for the product states @xmath106 correlations decay exponentially , opposite to what is expected in the superfluid phase , whose correlations should decay algebraically .
nevertheless , as we will see next , this family of states does estimate accurately the local properties of the optical lattice .
we will now compare the results for the ground state energy , the correlation functions , and the variance of the particle number provided by the two variational ansatz , ( [ mathieu ] ) and ( [ anglin ] ) , and the gutzwiller ansatz @xcite with those obtained by dmrg studies of the bose - hubbard model .
the dmrg , developed 1992 by white @xcite in the area of condensed matter theory , is a very powerful numerical tool to investigate static and dynamic properties of strongly correlated quasi - one - dimensional spin , fermionic or bosonic quantum systems .
the dmrg is an essentially quasi - exact numerical method .
the fundamental ideas stem from real space renormalization methods : the system size is grown iteratively while the ( exponentially diverging ) size of the hilbert space is kept constant by decimation .
hereby one tries to retain only that subset of states that is essential to describe the physical quantity under consideration . in dmrg
these are expectation values with respect to low - lying states ( `` target states '' ) , and in particular with respect to the ground state wave function . dmrg builds up the system linearly : at each growth step , suitable density matrices for the target states are derived that yield information on the relevance of hilbert space states .
building on this information , the states and operators are projected onto hilbert subspaces of fixed dimension @xmath2 containing the most relevant states .
@xmath2 is chosen to be small enough to be handled numerically , but large enough to obtain the desired accuracy ; numerical results can be extrapolated in @xmath2 to the exact limit of infinite @xmath2 in the thermodynamic limit .
however , results presented here have converged for the largest @xmath2 considered and no further extrapolation was necessary .
details on the dmrg method may for example be found in @xcite . in the case of the bose - hubbard model
the dmrg has been used to study properties of the system @xcite .
the results of dmrg agree very well with exact diagonalization results for small systems , with quantum monte - carlo simulations e.g. @xcite , and with 13th order perturbation theory @xcite .
we have used the dmrg to study the properties of the ground state of the bose - hubbard model on one - dimensional lattices with 128 sites , and commensurate fillings @xmath107 and @xmath108 . in figs . [ fig - dmrg](a - c ) we show the results for the mean energy per site @xmath109 , the nearest neighbor correlations , @xmath110 , and the variance @xmath59 of the density , calculated both with the dmrg and with the variational estimates developed above . as expected ,
there are no indications of the phase transitions in these quantities , neither in the variational results nor in the numerical solutions .
rather , an inflexion of the nearest neighbor correlation points out the location of the superfluid - insulator transition which lies roughly between @xmath111 and @xmath112 ( see @xcite and ref . therein ) .
the agreement of the two methods is fairly good above the phase transition and below it .
a more detailed comparison is provided in fig .
[ fig - gutz ] for the case @xmath113 . in this figure
we plot together results from the dmrg , the variational ansatz derived above , the quantum rotor model and the well - known gutzwiller ansatz .
the gutzwiller ansatz @xcite is a variational ansatz which reduces the wave function to a product of single - site wave functions , @xmath114 , where @xmath115 and @xmath116 are constants .
such a wavefunction can not be used in the one - dimensional mott insulator regime , because a perturbative study of the gutzwiller ansatz shows that the corrections of order @xmath117 are lost and all correlations become zero .
however this ansatz gives good results in the superfluid regime , where the long - range order is well described by @xmath118 , and we can use these results and those of the dmrg to assert the accuracy of our variational estimates . as fig .
[ fig - gutz ] shows , in the mott insulator regime , the dmrg results agrees perfectly with our variational theory for the bose - hubbard model and for the quantum rotor model .
close to the phase transition is the point at which the quantum rotor model no longer describes well the atoms in the optical lattice due to the growth of density fluctuations . at this point
we also observe a small disagreement between the dmrg and the coherent states , which is due to the growth of long range correlations and vanishes as we get deeper into the superfluid regime . in this section
we will show how to compute the corrections to the local energy , @xmath102 , from eq .
( [ var - non - h ] ) .
we basically need a method to compute expectation values of the operator @xmath119 around product states which have the form @xmath120 in which at most two contiguous vectors are affected by two single - connection operators .
for instance , this is the case of @xmath121 , where @xmath122 and @xmath123 are @xmath124 , @xmath125 or @xmath126 , and also of @xmath127 , where @xmath122 and @xmath123 are just the identity .
after some manipulations it is possible to write , for the optimal variational state @xmath128 , @xmath129 with the real matrices and vectors @xmath130^{-1/2},&i - j\geq-\bar{n}\\ 0,&i - j<-\bar{n}\end{array}\right.,\\ u_i & = & \delta_{i0},\\ i , j&\in&\mathbb{z},\;\alpha\in\{+,-,z\}.\end{aligned}\ ] ] we have used this technique to compute numerically the correction @xmath102 using different lattice sizes and found small or no differences for more than 30 sites . intuitively , this is because in the limit of large powers the matrices @xmath131 become projectors on the eigenvector with the largest eigenvalue .
using the same type of expansion we may compute other correlations @xmath132 for large values of @xmath133 and large lattices , the numerator will decay exponentially as @xmath134 , where @xmath135 is the largest eigenvalue of the matrix @xmath136 .
in this work we have studied analytically and numerically the properties of the ground state of an ensemble of bosonic atoms in an 1d optical lattice . for the study of the atomic ensemble we have used both the quantum rotor model and the bose - hubbard model . exploiting the fact that in these models there exists only nearest neighbor hopping and local interactions
, we have developed a variational wavefunction that may be used to easily estimate local properties , such as the energy per well , the nearest neigbor correlations and the fluctuations of the density . in the case of the quantum rotor model
we have verified our results with perturbative calculations around the strongly interacting regime , and with a spin wave approximation around the superfluid regime . in the case of the bose - hubbard model
we have compared the variational estimates with numerical results obtained using the dmrg technique for a maximum density of three atoms per well .
we have concluded that this procedure leads to fairly good estimates of local ground state properties of both hamiltonians , in both the superfluid and the insulator regime , the largest disagreement being localized around the phase transition . in future work
we forsee the application of these methods to time dependent problems .
we thank w. zwerger for fruitful discussions . j.
j. garca - ripoll and j. i. cirac thank the deutsche forschungsgemeinschaft ( sfb 631 ) and the kompetenznetzwerk quanteninformationsverarbeitung der bayerischen staatsregierung . c. k. and u. s.
thank the hess - preis of the dfg and the studienstiftung des deutschen volkes for financial support .
work at the university of innsbruck is supported by the austrian science foundation , eu networks and the institute for quantum information .
thanks the max planck institut fr quantenoptik for hospitality during his stay in garching , and thanks the humboldt foundation for support during this stay . | we develop a variational wave function for the ground state of a one - dimensional bosonic lattice gas .
the variational theory is initally developed for the quantum rotor model and later on extended to the bose - hubbard model .
this theory is compared with quasi - exact numerical results obtained by density matrix renormalization group ( dmrg ) studies and with results from other analytical approximations .
our approach accurately gives local properties for strong and weak interactions , and it also describes the crossover from the superfluid phase to the mott - insulator phase . |
neutron stars are the remnants of supernova explosions with masses 1@xmath0 , radii @xmath1 , and a temperature of the order of @xmath2 at birth , cooling within a few days to about @xmath3 by emitting neutrinos . in a conventional model for a neutron star ,
the star is composed of hadrons , predominantly degenerate neutrons with an admixture of protons , and degenerate electrons . in the stellar modeling ,
the structure of the star depends on the assumed equation of state ( eos ) , which contains a number of uncertainties associated with uncertainties in the strong force under conditions appropriate to a neutron star .
an important uncertainty concerns the true ground star of matter . in conventional models ,
hadrons are assumed to be the true ground state of the strong interaction .
however , it has been argued @xcite that ` strange matter ' is the true ground state of all matter .
if this is correct , the interior of neutron stars should be composed predominantly of @xmath4 quarks , plus leptons to ensure charge neutrality .
this led to the suggestion that there may be no neutron stars , and that all neutron - like stars are in fact strange stars @xcite .
more generally , the composition of neutron stars remains a source of speculation , with some of the possibilities being the presence of hyperons @xcite , a mixed phase of hyperons and quarks @xcite , a phase of deconfined quarks or pion and kaon condensates @xcite .
models in which the interior is assumed to be composed of strange matter are often called ` strange ' stars . however , because , as we show , the strangeness content depends on the model used to describe the quark matter , we prefer to describe any model in which the interior involves deconvolved quarks ( not bound in hyperons ) as ` quark ' stars . in the stellar modeling ,
the structure of the star depends on the assumed eos , which is different in each of these cases , and which depends on the nature of the strong interaction .
apart from the differences in the eos , an important distinction between quark stars and conventional neutron stars is that the quark stars are self - bound by the strong interaction , whereas neutron stars are bound by gravity .
this allows a quark star to rotate faster than would be possible for a neutron star .
further evidence in favor of quark stars is that some stars do indeed seem to rotate faster than what would be expected for a neutron star @xcite .
a quark star must have a thin layer on its surface dominated by the electrons which are necessary to enforce charge neutrality .
this layer could suspend a hadronic crust , which would not be in contact with the stellar core @xcite .
it has been suggested that the presence of a crust provides a natural explanation for pulsar glitches @xcite , which are sudden changes in the rotation period of the pulsar .
other authors have argued that quark stars should be ` bare ' @xcite , in the sense that any such crust would either not form or would be destroyed during the supernova explosion . recently , the characteristics of the radiation from hot , bare strange stars have been identified @xcite .
possible candidates for quark stars include compact star with gravitational masses around @xmath5 and radii of the order of @xmath6 : psr j1645 - 0317 ( psr b1642 - 03 ) , psr j1830 - 1059 ( psr b1828 - 11 ) @xcite , rx j185635 - 3754 @xcite , her x-1 @xcite , 4u 1728 - 34 @xcite , x - ray bursters gro j1744 - 28 @xcite and sax j1808.4 - 3658 @xcite .
there are several different models that describe quark matter . in @xcite
we investigated the properties of strange stars using two different models : the mit bag model @xcite , which is widely favored in the literature on strange stars , and the color - favor - locked phase ( cfl ) model @xcite , which allows the quarks near the fermi surface to form cooper pairs which condense and break the color gauge symmetry @xcite . at sufficiently high density
the favored phase is called cfl , in which quarks of all three colors and all three flavors are allowed to pair . in @xcite
we verified that quark stars have different maximum gravitational and baryonic masses , compared with neutron stars , and that the maximum masses are not significantly affected by the presence or absence of a crust .
moreover , with a simple prescription for the kepler frequency , which determines the spin rate of a neutron - like star , we obtained much higher values for the quark stars than for the neutron stars .
finally , in @xcite we used the two possible mass - to - radius ratio constraints available in the literature @xcite to distinguish between the assumed eoss .
concerning one of the constraints , specifically the interpretation of absorption features as atomic transition lines in @xcite is controversial : an alternative interpretation @xcite is that the aborption features are cyclotron lines , which imply no obvious constraint on the eos .
however , if the accepted interpretation is correct , then only the quark star eos is compatible with the constraint given in @xcite , thereby excluding most of the neutron star eoss .
the conclusion that all neutron stars must be quark stars has major implications , but it is necessarily valid only for either of the two eoss investigated in @xcite . in the present paper we focus on a different model for the strange matter , the nambu - jona - lasinio ( njl ) model @xcite .
our aim is to use this model to obtain the stellar properties and compare the results with the mit bag model @xcite .
it is important to distinguish between the eos during the short time period when neutrinos are still trapped in the star , and the eos after the neutrons escape .
the maximum entropy per baryon ( @xmath7 ) reached in the core of a new born star is about 2 ( in units of boltzmann s constant ) @xcite .
we then perform our calculations for @xmath8 ) , 1 and 2 since the entropy and not the temperature should be constant throughout the star @xcite . in section ii the formalism used is presented ; in section iii we give the results and make the relevant discussions and in section iv the conclusions are drawn .
in this section we summarize the main formulae for both models used in this paper .
we choose the njl model @xcite to describe the quark phase .
the su(3 ) version of the model includes most of symmetries of qcd , including chiral symmetry , and its breaking , which is essential in treating the lightest hadrons .
the njl model also includes a scalar - pseudoscalar interaction and the t hooft six fermion interaction that models the axial @xmath9 symmetry breaking .
the njl model assumes deconfined point - like quarks , and is not remornalizable , requiring regularization through a cutoff in three - momentum space .
the njl model is defined by the lagrangian density @xmath10\nonumber\\ & + & \
\,g_d\,\ , \{{\mbox{det}\,[\bar q_i\,(1+\gamma_5)\,q_j ] + \mbox{det}\,[\bar q_i\,(1-\gamma_5)\,q _ j]\ , } \},\label{1}\end{aligned}\ ] ] where @xmath11 are the quark fields and @xmath12 @xmath13 are the u(3 ) flavor matrices .
the model parameters are : @xmath14 , the current quark mass matrix ( @xmath15 ) , the coupling constants @xmath16 and @xmath17 and the cutoff in three - momentum space , @xmath18.the njl model is valid only for quark momenta smaller than the cut - off @xmath18 .
the set of parameters is chosen in order to fit the values in vacuum for the pion mass , the pion decay constant , the kaon mass and the quark condensates .
we consider the set of parameters @xcite
: @xmath19 , @xmath20
, @xmath21 , @xmath22 and @xmath23 which are fitted to the following properties : @xmath24 , @xmath25 , @xmath26 , @xmath27 , @xmath28 and @xmath29 . the thermodynamical potential density is given by @xmath30 , where the energy density is @xmath31 and the entropy density is @xmath32 + [ n_{i+}\rightarrow n_{i-}]\right\}. \label{5}\end{aligned}\ ] ] in the above expressions @xmath33 , @xmath34 is the temperature , @xmath35 ( @xmath36 ) is the chemical potential ( number ) of particles of type @xmath37 , and @xmath38 and @xmath39 are included in order to ensure @xmath40 in the vacuum .
this requirement fixes the density independent part of the eos .
the ground state of the system is described by the density matrix @xcite given by @xmath41 with @xmath42\ , \theta ( \lambda^2 -p^2),\label{2}\ ] ] where @xmath43 is the identity matrix , @xmath44 are the fermi distribution functions of the negative ( positive ) energy states , @xmath45^{-1 } , \quad i = u , d , s.$ ] in the last equation @xmath46 , @xmath47 is the constituent quark mass , @xmath48 .
the quark condensates and the quark densities are defined , for each of the flavors @xmath49 , respectively , as : @xmath50 @xmath51 minimizing the thermodynamical potential @xmath52 with respect to the constituent quark masses @xmath47 leads to three gap equations for the masses @xmath47 @xmath53 with cyclic permutations of @xmath54 .
we introduce an effective dynamical bag pressure @xcite , @xmath55 in terms of this quantity the energy density ( [ 4 ] ) takes the form @xmath56 where @xmath57 .
writing the energy density in terms of @xmath58 allows us to identify this contribution as a bag pressure and establish a relation with the mit bag model @xcite discussed in the next section .
the mit bag model @xcite has been extensively used to describe quark matter . in its simplest form ,
the quarks are considered to be free inside a bag and the thermodynamic properties are derived from the fermi gas model .
the energy density , the pressure and the quark @xmath59 density , respectively , are given by @xmath60 @xmath61 @xmath62 where @xmath63 stands for the number of colors , @xmath64 for the spin degeneracy , @xmath65 for the quark masses , @xmath66 represents the bag pressure and the distribution functions for the quarks and anti - quarks are the fermi distributions @xmath67})\ ; , \label{distf}\ ] ] with @xmath68 ( @xmath69 ) being the chemical potential for quarks ( anti - quarks ) of type @xmath59 and @xmath70 .
these equations apply at nonzero temperatures . for @xmath71
, there are no antiparticles , and the particle distribution functions become the usual step functions .
we use @xmath72 , @xmath73 and @xmath74 . if @xmath75 and @xmath76 are chosen as in the njl model , the behavior of the properties of interest are not altered , since they are more dependent on the bag pressure than on small differences in the quark masses . in a star with quark matter
we must impose both beta equilibrium and charge neutrality @xcite . in what follows
two scenarios are investigated , an early stage when there are trapped neutrinos in the interior of the star and a later stage , after the neutrinos escape ( deleptonization ) .
we first consider the later stage when entropy is maximum and neutrinos diffuse out .
the neutrino chemical potential is then zero . for @xmath77-equilibrium matter we must add the contribution of the leptons as free fermi gases ( electrons and muons ) to the energy and pressure .
the relations between the chemical potentials of the different particles are given by @xmath78 for charge neutrality we must impose @xmath79 for the electron and muon densities we have @xmath80 where the distribution functions for the leptons are given in eq .
( [ distf ] ) by substituting @xmath59 by @xmath81 , with @xmath82 the chemical potential for leptons of type @xmath81 . at @xmath71 ,
( [ rhol ] ) becomes @xmath83 the pressure for the leptons is @xmath84 in earlier stage , when the neutrinos are still trapped in the interior of the star , eq .
( [ qch ] ) is replaced by @xmath85 the lepton contribution is set to be @xmath86 @xcite .
no muons appear in this case .
in order to describe the crust of the quark stars we next use the well known eos calculated in @xcite for very low densities . in figure [ mu ] the strange quark mass as a function of the density
is shown for the njl models and the 6 cases discussed in the present work , i.e. , @xmath87 without neutrinos and with trapped neutrinos .
these data have a substantial influence on the results we obtain with the njl model .
chiral symmetry restoration occurs first for the @xmath88 and @xmath89 quarks and only at higher densities for the @xmath90 quarks .
this is clearly seen in figure [ mu ] : the mass of the @xmath90-quark starts to decrease only for @xmath91 for neutrino free ( trapped ) matter . in figure [ fig1 ]
we show the eos obtained respectively for the mit bag and for the njl models again for the 6 different cases mentioned above .
as shown in the upper panel in figure [ fig1 ] , the results are very similar and it is difficult to distinghish between them ; this may be attributed to the familiar result that for an extreme relativistic gas the pressure is one third of the energy density .
the njl model presents different behaviors at high densities .
the fact that chiral symmetry is restored at different densities for quarks with different masses explains the softer behavior of the eos for @xmath92 . at these densities
the onset of strangeness occurs , with the onset occurring at larger densities when neutrino trapping is included ( see fig .
the eos with trapped neutrinos is harder and the @xmath87 cases are similar , but distinghishable since higher temperatures correspond to harder eoss .
this behavior was already seen in @xcite , although in a different context . in figure [ fig2 ]
we show the temperature range described by both models for the cases @xmath93 and @xmath94 . for the mit model ,
the temperature for each different eos increases steadily , reaching about @xmath95 for @xmath94 and @xmath96 for @xmath93 at @xmath97 .
the njl model presents a different behavior : the temperature first oscillates before increasing .
the near - plateau around 46 @xmath98 coincides with the onset of strangeness opening of a new degree of freedom .
the maximum temperature for @xmath94 without neutrinos is higher than the one obtained with the mit model for the same energy density .
this difference is due to the fact that within the mit model the fractions of the different types of quarks are similar , unlike the njl which has a smaller fraction of @xmath90-quarks , which gives rise to a higher entropy .
if we fix the entropy the temperature must be lower within the mit model .
however , from table i we can see that within the njl the central energy densities of compact stars are much lower than the corresponding ones for the mit model . the temperatures are not shown in table i because they vary though the star ; the temperatures in the interior of the quark stars are then similar for the njl and the mit models , despite the significant difference in density . including trapped neutrinos lowers the temperature of the star .
this is due to the presence of a fixed fraction of leptons increasing the degrees of freedom , and then to keep the entropy per particle and the thermodynamic potential fixed , a lower temperature is required .
if neutrino trapping is not imposed the fraction of leptons in the star is very small . in figure [ fig3 ]
we plot the strangeness content in each situation described in the text .
as already referred , within the njl the onset of strangeness occurs for @xmath99 depending on the temperature and the neutrino content .
a different situation applies for the mit bag : strangeness is present from lower densities and at @xmath99 the strangeness fraction is almost at its maximum value in the case without neutrinos , @xmath100 .
the presence of leptons in the case of trapped neutrinos with a fixed lepton fraction lowers the strangeness content due to the electric charge conservation restrictions . in figure [ fig4 ]
we display , for both the mit and the njl models , the electron neutrino fraction present in the interior of the stars when trapped neutrinos are included . with the mit model
the neutrino content is practically the same for the three entropies considered .
it increases a little for lower densities and then reaches its maximum value , for the same reasons discussed above , i.e. , neutrinos and electrons together are fixed to a certain fraction and electrons compensate for charge neutrality in a system where the three quarks have equal number densities .
the njl model shows a slight decrease in the neutrino fraction until the onset of strangeness .
this effect is more pronounced for @xmath101 when chiral symmetry restoration has still not occured for the @xmath88 and @xmath89 quarks .
a slight increase in the neutrino fraction occurs with the increase of the entropy , mainly at low densities .
the neutrino fraction never reaches 14@xmath102 within the njl picture , but it amounts to more than 15@xmath102 within the mit model .
we see a similar neutrino fraction for these models in @xcite .
the difference in neutrino content is again due to the larger fraction of strange quarks in the mit description and therefore a smaller number density of electrons required by the electric charge neutrality .
a smaller fraction of electrons implies a larger fraction of neutrinos for a fixed lepton fraction
. nevertheless , if the cfl model were used for the quark matter with trapping , a much higher neutrino fraction , of the order of 35@xmath102 , would be possible @xcite . given the eos , the next step is to solve the tolman - oppenheimer - volkoff equations @xcite .
table i lists the main properties of the stars , including their maximum gravitational and baryonic masses , their radii and central energy densities , for all the eoss considered in the present work . as expected , quark stars have smaller maximum masses with smaller radii @xcite and higher central densities .
let us first analyze the results obtained with the mit model and then outline the main differences that appear with the njl model .
for quark stars either without or with neutrinos , the maximum gravitational masses and radii are always around 1.231.24 @xmath103 and 6.7@xmath104 , respectively , and the energy density is around @xmath105 .
the inclusion of trapped neutrinos and a crust , have practically no effect on the numerical results .
the njl model leads to significantly different results .
the maximum masses and radii are systematically larger and , as a consequence , the central energy densities are lower than for the mit bag model . in fact , the results do not depend strongly on the entropy , as already pointed out in @xcite . in the njl , for a bare star without neutrinos , the maximum masses and radii decrease and the central energy density increases as the entropy increases . for a star with a crust , the gravitational and baryonic masses decrease but
the radii oscillate as the entropy increases . in fig .
[ fig5 ] we show for @xmath106 , and for both models the mass - radius plot of the family of stars obtained for the neutrino free and the neutrino trapped situations .
it is clear that the njl predicts the possibility of existence of more massive and with larger radius stars . for a fixed radius smaller than the larger accepted by the mit model
the njl gives stars with much smaller gravitation masses .
conclusions for finite entropies are the same . in a previous work @xcite
we see that the mit results are sensitive to the @xmath66 parameter . with a lower value of this parameter ,
results similar to the ones obtained with the njl model are found . the same is possible when the cfl model is used . in @xcite
we fixed the gap parameter of the cfl model as @xmath107 and showed that the properties of the star depend on the @xmath66 parameter , but the properties are also known to depend on the gap parameter @xcite .
it is not clear that quark stars really exist , and if the do exist in principle , how they would be formed .
they form directly in a supernova explosion , or as a consequence of the hadron - quark deconfinement phase transition in stellar compact stars @xcite . in the former case
there is neutrino trapping during the first seconds of the stars life and the entropy is approximatly constant throughout the star @xcite . in this case one
concludes from our results that there can be no late blackhole formation .
this event is only possible if the baryonic mass of a stable star at finite temperature is larger than the largest baryonic mass of a stable cold star @xcite .
from table i one can see that the maximum baryonic masses at fixed entropies are always lower than the corresponding values at zero temperature .
this kind of effect could occur within hybrid stars @xcite . | lately , it has been suggested that strange ( quark ) stars can be responsible for glitches and other observational features of pulsars . some discussions on whether quark stars , if really exist , are bare or crusted are also a source of controversy in the recent literature . in the present work we use the nambu - jona - lasinio model , known to incorporate chiral symmetry , necessarily present in the qcd formalism , in order to describe quark star properties .
we compare our results for the stars and the features of the model with the much simpler model normally used in the description of strange stars , namely the mit bag model .
we also investigate the differences in the stellar properties which arise due to the presence of the crust .
we show that the njl model produces results which are somewhat different as compared with the mit model .
pacs number(s ) : 26.60.+c,24.10.jv , 21.65.+f,95.30.tg |
while the fundamental equations of motion , both in classical and quantum mechanics , are invariant with respect to time reversal , the world around us is full of irreversible processes . on a microscopic level , it is enough to recall spontaneous decays of particles , nuclei , inelastic scattering processes , etc . and , of course , an irreversible process _
par excellence _ is the wave packet reduction which is the core of copenhagen description of a measuring process performed on a quantum system . a description of such processes in quantum mechanics is typically associated with enlarging the state hilbert space , conventionally referred to as coupling the system to a heat bath @xcite .
as a rule , it is assumed that ( i ) the bath is a system with infinite number of degrees of freedom , ( ii ) the bath hamiltonian has a continuous spectrum , and ( iii ) the presence ( or absence ) of irreversible modes is determined by the energies involved rather than the coupling strength . while this all is true in many cases
, situations may exist where the system has neither of the listed properties .
this was the motivation which led uzy smilansky to formulation a simple model @xcite of a quantum graph coupled to a heat bath consisting of a single harmonic oscillator which exhibits an irreversible behavior . the easiest way to describe
the model is to phrase it in pde terms , then its hamiltonian is a schrdinger operator , @xmath0 on @xmath1 , with a singular interaction supported by the line @xmath2 the strength of which depends on the coordinate @xmath3 .
the system governed by the hamiltonian undergoes an abrupt spectral transition at the critical value @xmath4 ; while for @xmath5 the hamiltonian is positive above this value its spectrum covers the whole real axis .
the mechanism of this effect is easy to understand .
the first and the last term in represent hamiltonian of a one - dimensional system with a point interaction which has a single eigenvalue equal to @xmath6 @xcite .
this negative contribution to the energy competes with the oscillator potential with the balance flipping exactly at the said critical value of @xmath7 .
the model was subsequently an object of investigations , generalizations and modifications in a series of mathematically rigorous papers .
the discrete spectrum in the subcritical case and its behavior as @xmath8 were analyzed in @xcite , the absolute continuity of the other spectral component was established in @xcite , an extension to the situation when the ` heat bath ' consists of two or more oscillators was discussed in @xcite .
other modifications consisted of replacing the oscillator by a potential well of a different shape @xcite or by replacing the line by a more general graph @xcite corresponding to the original smilansky proposal .
it is also possible to have the motion in the @xmath9 direction restricted to an interval with periodic boundary conditions @xcite ; in the first named of these papers time evolution of wave packets is investigated confirming the idea that spectral change in the supercritical case corresponds to the possibility of ` escape to infinity ' .
let us also mention a modification in which the singular interaction is replaced by a regular potential channel which gets deeper and more narrow as the distance from a fixed point increases @xcite . in connection with the last mentioned system
let us point out that similar parameter dependent spectral transitions can be also observed in other systems , an example being the one described by the hamiltonian @xmath10 on @xmath11 @xcite . in this case
the change is even more dramatic , from the spectrum which is below bounded and purely discrete to the whole real line , the transition occurring at the critical value @xmath12 , the ground state eigenvalue of the appropriate ( an)harmonic oscillator . to our knowledge ,
spectral transitions of this type were for the first time noted in @xcite .
our object of interest in the present paper is the original smilansky hamiltonian . despite the fact that many of its properties were demonstrated in the papers mentioned above
, there is still a room for new insights .
the first topic here concerns the discrete spectrum of @xmath13 and comes from the observation that it is relatively easy to find the eigenvalues and eigenfunctions numerically turning the task into a matrix problem of a sufficiently large dimension .
another question to be discussed here concerns _
resonances _ in the system described by the hamiltonian . while the absolute continuity of the spectrum above the threshold and the existence of the wave operators were established in @xcite , the possible resonance character of the scattering seems to have escaped attention .
in our recent paper @xcite we have demonstrated how one can find poles of the analytically continued resolvent and derived asymptotic expansions for them , here we focus on the physical meaning of the resonances and illustrate their behavior by numerical results .
to make the paper self - contained we summarize first the known facts about the operator that are indicated briefly in the introduction .
it is obvious that we can consider the positive values of the coupling constant @xmath7 as the opposite case is obtained via a mirror transformation with respect to the axis @xmath2 . _
( a ) spectral transition : _ in the subcritical case , @xmath14 , the particle remains localized in the vicinity of the @xmath9 axis , while for @xmath15 it can according to @xcite escape to infinity along the singular ` channel ' in the @xmath3 direction . in spectral terms
, this corresponds to the switch from a positive spectrum of @xmath13 to a below unbounded one which occurs at @xmath16 .
_ ( b ) the continuous spectrum _ of @xmath13 covers the interval @xmath17 for @xmath14 , the positive semi - axis for @xmath18 , and the whole real axis if @xmath15 .
moreover , this spectrum is in fact purely absolutely continuous ( * ? ? ?
vii.2 ) , in particular , the operator is purely absolutely continuous in the supercritical case when @xmath19 . _
( c ) eigenvalue existence and location : _ for any @xmath20 the discrete spectrum of @xmath13 is nonempty , simple , and finite , being contained in @xmath21 , i.e. the subcritical operator is positive .
the eigenvalues are decreasing functions of the coupling constant @xmath7 . on the other hand , there are no eigenvalues @xmath22 , and in the supercritical case , @xmath15 , the point spectrum of @xmath13 is empty , in other words , there are no eigenvalues embedded in the continuum . _
( d ) near critical asymptotics : _ the number of eigenvalues in the subcritical regime increases with the coupling constant and explodes as it approaches the critical value @xmath16 behaving asymptotically as @xcite @xmath23 where @xmath24 conventionally denotes the number of eigenvalues of the operator @xmath25 below @xmath26 with the multiplicity taken into account . _
( e ) weak - coupling asymptotics : _ for small enough @xmath7 there is a single eigenvalue which behaves as @xmath27 as @xmath28 @xcite .
in fact , due to the mirror symmetry of the spectrum with respect to @xmath7 the error term can be replaced by @xmath29 .
to find the discrete spectrum we note first that the solutions to the eigenvalue problem can be expanded using the ` transverse ' base naturally spanned by the functions @xmath30 corresponding to the oscillator eigenvalues @xmath31 ; here @xmath32 are hermite polynomials with conventional normalization .
furthermore , the task can be simplified by noting that the problem has a mirror symmetry w.r.t . which allows us to decompose @xmath13 into the trivial odd part @xmath33 and the even part @xmath34 which is equivalent to the operator on @xmath35 , where @xmath36 , with the symbol and the domain consisting of @xmath37 functions satisfying the boundary conditions @xmath38 substituting the ansatz @xmath39 with @xmath40 into and using orthonormality of the basis , we get for solution with the energy @xmath41 the equation @xmath42 where @xmath43 is the coefficient vector and @xmath44 is an operator in @xmath45 , in other words , an infinite matrix with the elements @xmath46 note that the matrix @xmath44 is in fact tridiagonal because @xmath47 in other words , @xmath44 is a jacobi matrix .
note that using a general sequence @xmath48 instead of @xmath49 in the ansatz one can reformulate the problem as an analogue of the birman - schwinger principle as first proposed by m. solomyak in @xcite , see a detailed discussion in @xcite .
the same approach works also for some modifications of the original smilansky model mentioned in the introduction where , however , the structure of the appropriate matrix problem is different and may be more complicated . to find the values of @xmath41 for which the secular equation is satisfied for a given @xmath7 we solve a truncated matrix problem increasing the cut - off until a numerical stability is reached .
the convergence is slower the closer is the value of @xmath41 to the threshold value @xmath50 .
we present results for the sizes up to @xmath51 . as functions of @xmath7,width=453 ]
the results are plotted below in a series of figures .
they are , of course , in agreement with the theoretical predictions listed in the previous section , but they provide additional insights .
to begin with , fig .
[ ev_plot ] shows eigenvalues of @xmath13 as functions of the coupling constant .
we see that in the most part of the subcritical region , more than @xmath52 of it , the hamiltonian has a single eigenstate .
the second eigenvalue appears only at @xmath53 , the next thresholds are @xmath54 .
this does not give a good picture of the spectral behavior near the critical values .
instead we can plot the eigenvalue as functions of @xmath7 using a logarithmic scale .
the result is shown in fig .
[ log_plot ] .
we see that their number indeed grows as we approach the transition point filling the allowed interval so that @xmath55 , but that the discrete spectrum is asymptotically equidistantly distributed .
the same result allows us to illustrate the solomyak asymptotic formula : it is obvious from fig .
[ solombound ] that the leading term of the expansion gives a good approximation as long as @xmath56 .
the opposite asymptotical situation is the weak coupling . in fig .
[ weakfig ] we plot the ground - state eigenvalue as a function of @xmath7 . in the double logarithmic scale the graph is practically indistinguishable from a straight line which allows us to read out the power of the leading term as well as the coefficient @xmath57 appearing in the leading term of . since , in contrast to points ( a)(d ) of the previous section the weak - coupling asymptotics is a new observation it is worth to show a way how it can be derived . to this purpose
we introduce a symbol for the energy gap @xmath58 and denote by @xmath59 the normalized eigenfunction of @xmath44 given by corresponding to zero eigenvalue . without loss of generality we may assume that @xmath60 is non - negative .
writing the lines of the corresponding infinite system of equation explicitly , we get @xmath61 & \frac{\sqrt{k}\lambda}{2\sqrt{2}}\,c^\lambda_{k-1 } + \sqrt{k + \mu_\lambda}\,c^\lambda_k + \frac{\sqrt{k+1}\lambda}{2\sqrt{2}}\,c^\lambda_{k+1 } = 0\ , , \qquad k\ge 1\ , .
\end{split}\ ] ] using elementary inequalities @xmath62 and that @xmath63 , we get from the second relation in the estimate @xmath64 which implies @xmath65 and combining it with the normalization assumption , @xmath66 , we arrive at @xmath67 using @xmath66 once more , we conclude that @xmath68 then inequality together with the normalization yield @xmath69 . in combination with and the second relation of for @xmath70
this implies further @xmath71 using now and we obtain from the first relation in that @xmath72 which is nothing else but the formula .
in the counterpart to this paper @xcite we give a modified version of this argument which allows , at least in principle , to compute higher terms of this asymptotic expansion . for @xmath73,width=453 ] before proceeding further , let us also look at the eigenfunctions . in fig .
[ eigenfig ] we show a contour plot of the first six eigenfunctions of @xmath13 for @xmath73 , in other words , @xmath74 . in the side plots we feature restrictions of the eigenfunctions on the @xmath3-axis .
as expected the eigenfunctions are concentrated in the vicinity of the ` escape channel ' , i.e. the negative part of the @xmath3-axis .
the plots also show that they obey the usual restriction , commonly referred to as courant nodal domain theorem ; the inequality becomes sharp from the third eigenfunction .
the pattern is interesting , the odd eigenfunctions ( except the ground state ) have closed nodal lines only , the even ones have an unbounded nodal line . one wonders whether this property holds for the discrete spectrum generally .
our second main point is to show that smilansky model also exhibits interesting resonance behavior .
having said that , one has to explain first what is meant by a resonance .
there are different definitions , the two main being the _ resolvent resonances _ associated with poles in the analytic continuation of the resolvent over the cut(s ) corresponding to the continuous spectrum , and _ scattering resonances _ associated with singularities of the scattering matrix .
while these two can often be identified , this property has to be checked in each particular case ; recall that the former are determined by a single operator , the latter come from comparison of the free and interacting dynamics .
from the differential equation point of view finding poles of the analytically continued resolvent means to solve the eigenvalue equation for looking for a complex ` eigenvalue ' corresponding to an ` eigenfunction ' which does not belong to @xmath75 being exponentially increasing with @xmath76 . modifying the argument of the previous section which led to the infinite matrix problem the task is again reduced to the equation . a subtle point in this case , as with other ` guiding channel ' systems with infinite number of transverse modes @xcite producing infinite number of cuts in the continuous spectrum , is that the riemann surface of energy has infinite number of sheets , in fact an uncountable one .
we are interested in those of them neighboring with the physical sheet which in the present context means that looking for resonances on the @xmath77-th sheet we have to flip sign of the first @xmath78 signs of the square roots in ; for a more thorough discussion of this problem we refer the reader to @xcite .
let us turn to scattering resonances .
suppose that the incident wave comes in the @xmath79-th channel from the left with the energy @xmath80 .
we use the ansatz @xmath81 \sum_{n=0}^\infty t_{mn}\ , \mathrm{e}^{-ip_nx } \psi_n(y ) , & \quad x > 0 .
\end{cases}\ ] ] where we have denoted @xmath82 .
let us note that the on - shell scattering matrix with the elements has the size @xmath83 where @xmath84 $ ] , the other ` reflection ' and ` transmission ' amplitudes correspond to the evanescent modes present in the scattering solution .
the sign choice of the square roots signifying the riemann sheet choice is the same as in the previous case .
it is straightforward to compute from here the boundary values @xmath85 and @xmath86 . the continuity requirement at @xmath2 together with the orthonormality of the basis @xmath87 yields @xmath88 furthermore , substituting the boundary values from the ansatz into @xmath89 and integrating the obtained expression with @xmath90 we obtain @xmath91 the requirement of solvability of this system leads to the same condition as before which allows us to conclude that the _ resolvent and scattering resonances in smilansky model coincide_. the next question concerns the existence of resonances and their behavior as the coupling constant @xmath7 changes .
there are some standard mechanisms producing resonances .
one is a perturbation of embedded eigenvalues of the hamiltonian , however , we have seen that in the present case there are no such eigenvalues .
another possibility comes from perturbation of the singularities associated with the thresholds of the transverse channels , i.e. the branching points of the energy surface cuts the weak - coupling analysis follows the route as for the discrete spectrum and shows that for small @xmath7 a resonance pole splits of each threshold according to the asymptotic expansion .
this indeed leads to emergence of complex poles in the vicinity of thresholds which behave as @xmath92 as one can readily check repeating the argument of the previous section , in fact , the weak coupling of the bound state is included if we put @xmath93 in .
for @xmath94 we get genuine complex poles the distance for the corresponding threshold is proportional to @xmath95 and the trajectory asymptote is the ` steeper ' the larger @xmath79 is
. a more detailed discussion of the weak - coupling behavior of these resonances and also of resonances lying on the sheets that are not adjacent to the physical sheet can be again found in @xcite
. runs through the interval @xmath96 .
resonances that do not depart from thresholds are marked by dashed lines , width=377 ] however , this tells us nothing about the pole behavior for larger @xmath7 . to get a more complete picture
, we again analyze the problem numerically .
as described above , we look for complex values of @xmath41 for which the secular equation is satisfied for a given @xmath7 . now
, however , @xmath44 is modified by flipping sign of the first @xmath78 signs of the square roots .
then , a truncated matrix problem ia solved , tuning the cut - off until a numerical stability is reached .
the results are shown in fig .
[ resofig ] where we demonstrate resonances splitting from the lowest thresholds , from the second to the fifth one together with their weak - coupling asymptotes according to .
we see that the behavior beyond the weak - coupling regime is different , the pole trajectories return to the real axis and they may even end up as an isolated eigenvalue .
this is not surprising , however , since a similar behavior is known from other resonance models @xcite being first observed in the classical lee - friedrichs model @xcite .
what is even more interesting , the numerical analysis shows that the system has other resonances that do not emerge from thresholds and appear when the coupling surpasses a critical value . in fig .
[ resofig ] we show two such trajectories located at the second and third sheet of the energy surfaces and emerging at @xmath97 and @xmath98 , respectively .
one can conjecture naturally that these are not the only ones and that the system governed by smilansky hamiltonian has a rich family of resonances of different kinds which deserve a separate investigation .
this research was supported by the czech science foundation ( gar ) within the project 14 - 06818s .
99 s. albeverio , f. gesztesy , r. hegh - krohn , h. holden : _ solvable models in quantum mechanics _ ,
2nd edition with an appendix by p. exner , ams chelsea publishing , providence , r.i . , 2005 .
d. barseghyan , p. exner : a regular version of smilansky model , _ j. math
* 55 * ( 2014 ) , 042104 .
d. barseghyan , p. exner , a. khrabustovskyi , m. tater : spectral analysis of a class of schrdinger operators exhibiting a parameter - dependent spectral transition , _ j. phys .
a : math . theor . _ * 49 * ( 2016 ) , 165302 .
davies : _ open quantum systems _ , academic press , london 1976 .
evans , m. solomyak : smilansky s model of irreversible quantum graphs .
i : the absolutely continuous spectrum , _ j. phys .
a : math . gen . _
* 38 * ( 2005 ) , 46114627 .
evans , m. solomyak : smilansky s model of irreversible quantum graphs .
ii : the point spectrum , _ j. phys . a : math .
* 38 * ( 2005 ) , 76617675 .
exner : _ open quantum systems and feynman integrals _ ; fundamental theories of physics , vol . 6 , d. reidel , dordrecht 1985 .
exner : solvable models of resonances and decays , proceedings of the conference `` mathematical physics , spectral theory and stochastic analysis '' ( goslar 2011 ; m. demuth , w. kirsch , eds . ) , birkhuser , basel 2013 ; pp . 165227 .
exner , h. kovak : _ quantum waveguides _ ,
springer , cham 2015 .
p. exner , v. lotoreichik , m. tater : on resonances and bound states of smilansky hamiltonian , _ nanosystems : phys . chem . math . _ ( 2016 ) , to appear ; ` arxiv : 1607.00540 [ math - ph ] ` i. guarneri : irreversible behaviour and collapse of wave packets in a quantum system with point interactions , _ j. phys . a : math .
* 44 * ( 2011 ) , 485304 g. hhler : ber die exponentialnherung beim teilchenzerfall , _ zs . phys . _
* 152 * ( 1958 ) , 546565 .
s. naboko , m. solomyak : on the absolutely continuous spectrum in a model of an irreversible quantum graph , _ proc .
* 92 * ( 2006 ) , 251272 .
m. reed , b. simon : _ methods of modern mathematical physics .
i : functional analysis .
_ , academic press , new york 1980 .
g. rozenblum , m. solomyak : on a family of differential operators with the coupling parameter in the boundary condition ,
_ j. comput .
_ * 208 * ( 2007 ) , 5771 .
u. smilansky : irreversible quantum graphs , _ waves in random media _ * 14 * ( 2004 ) , s143153 .
m. solomyak : on the discrete spectrum of a family of differential operators , _ funct .
* 38 * ( 2004 ) , 217223 .
m. solomyak : on a mathematical model of irreversible quantum graphs , _ st . petersbg .
j. _ * 17 * ( 2006 ) , 835864 .
m. solomyak : on the limiting behaviour of the spectra of a family of differential operators , _ j. phys
. a : math . gen . _ * 39 * ( 2006 ) , 1047710489 .
m. znojil : quantum exotic : a repulsive and bottomless confining potential , _ j. phys .
a : math . gen . _ * 31 * ( 1998 ) , 33493355 . | we analyze the hamiltonian proposed by smilansky to describe irreversible dynamics in quantum graphs and studied further by solomyak and others .
we derive a weak - coupling asymptotics of the ground state and add new insights by finding the discrete spectrum numerically .
furthermore , we show that the model has a rich resonance structure . ` smilansky model , discrete spectrum , weak coupling , resonances 35j10 , 35p15 , 65n25 , 81q15 , 81q35 ` |
gamma ray bursts ( grbs ) are the brightest cosmological sources in the universe , with isotropic luminosities up to 10@xmath4 and a redshift distribution extending at least up to @xmath5 @xmath66.3 ( e.g. , tagliaferri et al . 2005 ) .
thus , these sources may be interesting for cosmological studies , if one can use them to provide measurements of the cosmological parameters independently of other methods , like the cmb ( e.g. , de bernardis et al . 2000 ; spergel et al .
2003 ; dunkley et al . 2008
; komatsu et al .
2008 ) , type ia sn ( e.g. , perlmutter et al . 1998 ; perlmutter et al .
1999 ; riess et al . 1998 ; riess et al . 2004 ; astier et al .
2006 ) , baryon acoustic oscillations ( bao ) ( e.g. , eisenstein et al .
2005 ; tegmark et al . 2006 ; percival et al .
2007 ) , galaxy clusters ( e.g. , rosati , borgani & norman 2002 ; voit 2005 ) .
however , grbs are not standard candles , given that their luminosities span several orders of magnitude under the assumption of both isotropic and collimated emission ( e.g. , ghirlanda et al .
in the recent years , several attempts to `` standardize '' grb have been made , mainly on the basis of correlations involving : a ) a grb intensity indicator , like the isotropic equivalent radiated energy ( ) or the isotropic
equivalent peak luminosity ( ) ; b ) the photon energy at which the time averaged spectrum peaks ( `` peak energy '' , ) ; c ) other observables , like the break time of the afterglow light curve , t@xmath7 , or the `` high signal time scale '' t@xmath8 @xcite .
finally , @xcite used different correlations ( spectrum energy , time lag luminosity , peak luminosity variability ) to construct a grb hubble diagram and constrain the cosmological parameters .
these analyses have provided useful constraints on and consistent with those derived from type ia sne , see , e.g. , @xcite for a review . however , the use of these correlations for cosmology is controversial .
for example , because of the lack of low redshift grbs they can not be directly calibrated . on the other hand the calibration of a spectrum energy correlation using sne - ia as calibrators @xcite is an interesting attempt to use grbs as cosmological probes , but it may look suspicious that this method gives very similar results to those obtained via sne . thus , particular care and sophisticated statistical methodologies have to be adopted in order to avoid circularity problems when constructing a grb hubble diagram .
in addition , recent analyses based on updated samples of grbs showed that the dispersion of three parameter correlations could be significantly larger than thought before @xcite . in this article
we investigate the possibility of constraining the cosmological parameters from the correlation .
this correlation was initially discovered on a small sample of grbs with known redshift @xcite and confirmed afterwards by swift observations @xcite .
although it was the first `` spectrum energy '' correlation discovered for grbs ( and the most firmly established one , to date ) it was never used for cosmology purposes , because of its significant `` extrinsic '' scatter ( i.e. , a scatter in excess to the `` intrinsic '' poissonian fluctuations of the data ) .
however , the conspicuous increasing of grb discoveries combined with the fact that correlation needs only two parameters that are directly inferred from observations ( this fact minimizes the effects of systematics and increases the number of grbs that can be used , by a factor @xmath3 ) makes the use of this correlation an interesting tool for cosmology .
our study was motivated by our finding that , in the assumption of a flat universe , the @xmath9 obtained by fitting the correlation with a power - law varies with the value of assumed to compute , has its minimum value for @xmath60.25 , very close the value obtained with sne - ia approach . in this paper
we show indeed that a small fraction of the extrinsic scatter is due to to the choice of cosmological parameters , thus allowing us to constrain and , to a smaller extent , . [ cols="<,<,<,<,<,<,^ " , ]
we have used an updated sample of 70 grbs aimed at deriving the cosmological parameters and from the correlation of grbs . with respect to previous attempts
mainly based on the and t@xmath8 correlations , the use of correlation has the advantages of requiring only two parameters , both directly inferred from observations .
this fact allows for the use of a richer sample of grbs ( e.g. , a factor 3 - 4 larger than used in the correlation ) , and a reduction of systematics , with respect to three
parameters spectrum energy correlations .
our method consists of studying the dependence of the dispersion of the correlation on cosmological parameters by adopting a maximum likelihood method , which allowed us to parameterise and quantify correctly the extrinsic ( i.e. non poissonian ) scatter component . from our analysis
, a number of significant results do emerge : \a ) a simple power
law fit shows a clear trend of the dispersion , with a minimum at = 0.20.4 ( for a flat universe ) . by refining the fit with the mlm
, we found that both the extrinsic variance and the slope of the correlation show a significant dependence on cosmology . from the study of the we derived = 0.040.40 at 68% c.l and 0.02 - 0.68% at 90% confidence levels ; \c ) if we release the assumption of a flat universe , we still find evidence for a value of @xmath11 ( 0.040.50 at 68% c.l . ) and a weak dependence of the dispersion of the correlation on @xmath12 with an upper limit of 1.15 ( 90% c.l . ) ; \d ) our study does not make assumptions on the correlation or make use of independent calibrators to set the `` zero point '' of the relation , therefore our approach does not suffer from circularity and provides independent evidence for the existence of a gravitationally repulsive energy component ( `` dark energy '' ) which accounts for a large fraction of the energy density of the universe ; \f ) we have simulated the impact of ongoing / planned grb experiments ( e.g. , + glast ) on the future and measurements and shown that the uncertainties can be decreased by almost an order of magnitude with respect to those obtained with the current grb sample .
amati l. et al . , 2002 ,
a&a , 390 , 81 amati l. , 2006 , mnras , 372 , 233 amati l. et al . , 2007 , a&a , 463 , 913 astier p. et al .
2006 , a&a , 447 , 31 bendat j.s .
, piersol a.g .
, 2000 , `` random data : analysis and measurement procedure
3rd ed . '' , wiley series in probability and statistics , john wiley & sons , inc .
, new york ( usa ) bevington p.r . ,
1969 , `` data reduction and error analysis for the physical sciences '' , mcgraw hill , new york ( usa ) butler n.r .
, kocevski d. , bloom j.s .
, curtis j.l . , 2007 , apj , 671 , 656 cabrera i.j .
et al . , 2007 , mnras , 382 , 342 campana s. et al . , 2006 ,
nature , 442 , 1008 campana s. et al .
, 2007 , a&a , 472 , 395 cenko s.b . , kasliwal m. , harrison f.a .
, 2006 , apj , 652 , 490 chevallier m. , polarski d. , 2001 , int . j. mod .
phys . , d10 , 213 dagostini g. , 2005 , physics/0511182 de bernardis p.
et al . , 2003 , nature , 404 , 955 dunkley j. et al .
, 2008 , apjs , submitted ( arxiv:0803.0586 ) eisenstein d.j .
et al . , 2005 ,
apj , 633 , 560 firmani c. et al .
, 2006 , mnras , 370 , 185 ghirlanda g. , ghisellini g. , lazzati d. , 2004 , apj , 616 , 331 ghirlanda g. , ghisellini g. , firmani c. , 2006 , new journal of physics , 8 , 123 ghirlanda g. , nava l. , ghisellini g. , firmani c. , 2007 , a&a , 466 , 127 ghirlanda g. , nava l. , ghisellini g. , firmani c. , cabrera j.i . , 2008 ,
mnras , 387 , 319 golenetskii s. et al.,2006 , gcn 4539 golenetskii s. et al.,2006 , gcn 4989 golenetskii s. et al.,2006 , gcn 5460 golenetskii s. et al . , 2007 , gcn 6049 golenetskii s. et al . , 2007 , gcn 6879 golenetskii s. et al . , 2007 , gcn 6960 golenetskii s. et al . , 2007 , gcn 7114 golenetskii s. et al . , 2008 ,
gcn 7482 golenetskii s. et al . , 2008 , gcn 7487 golenetskii s. et al . , 2008 ,
gcn 7589 guidorzi c. et al .
, 2006 , mnras , 371 , 843 kodama y. et al . , 2008 , mnras , in press ( arxiv:0802.3428 ) komatsu j. et al . , 2008 , apjs , submitted ( arxiv:0803.0547 ) krimm h. , et al . , 2006 , aip proc . , 836 , 145 landi r. , amati l. , guidorzi c. , frontera f. , montanari e. , 2006 , nuovo cimento b , 121 , 1503 liang n. , xiao w.k . , liu y. , zhang s.n . , 2008 ,
apjl , 685 , 354 linder e.v .
, utherer d. , 2005 , phrvd , 72 , 3509 mszros p. , 2006
, prog.phys , 69 , 2269 mundell c.g .
et al . 2007 , apj , 660 , 489 percival w.j .
et al . 2007 ,
mnras , 381 , 1053 perley c.g .
2008 , apj , 672 , 449 perlmutter s. et al .
1998 , nature , 391 , 51 perlmutter s. et al . 1999 ,
apj , 517 , 565 reichart d.e . , 2001 , apj , 553 , 235 riess a.g .
et al . , 1998 , apj , 116 , 1009 riess a.g .
et al . , 2004 ,
apj , 607 , 665 rosati p. , borgani s. , norman c. , 2002 , ara&a , 40 , 539 rossi et al . , 2008 , mnras , 388 , 1284 sakamoto t. et al . , 2005 ,
apj , 629 , 311 sakamoto t. et al . , 2008 , apjs , 175 , 169 sakamoto t. et al . , 2008 ,
apj , 679 , 570 salvaterra r. , campana s. , chincarini g. , covino s. , tagliaferri g , 2007 , mnras , 385 , 189 schaefer b.e . , 2007 , apj , 660 , 16 spergel d.n .
et al . , 2003 , apjs , 148 , 175 stratta g. et al . , 2006 ,
a&a , 461 , 485 tagliaferri g. et al . , 2005 , a&a , 443 , l1 tegmark m. et al . , 2006 ,
phrvd , 74 , 3507 thompson t. et al .
, 2007 , apj , 602 , 875 ulanov m.v .
et al . , 2005 , nuovo cimento c , 28 , 351 voit g.m .
et al . , 2005 , rev.mod.phys , 77 , 207 zhang , b. , mszros , p. , 2002 ,
apj , 581 , 1236 | we have used the correlation of grbs to measure the cosmological parameter . by adopting a maximum likelihood approach which allows us to correctly quantify the extrinsic ( i.e. non poissonian ) scatter of the correlation , we constrain ( for a flat universe ) to 0.040.40 ( 68% confidence level ) , with a best fit value of @xmath0 , and exclude = 1 at @xmath1 99.9% confidence level .
if we release the assumption of a flat universe , we still find evidence for a low value of ( 0.040.50 at 68% confidence level ) and a weak dependence of the dispersion of the correlation on ( with an upper limit of @xmath2 at 90% confidence level ) .
our approach makes no assumptions on the correlation and it does not use other calibrators to set the zero point of the relation , therefore our treatment of the data is not affected by circularity and the results are independent of those derived via type ia sne ( or other cosmological probes ) . unlike other multi - parameters correlations , our analysis grounds on only two parameters , then including a larger number ( a factor @xmath3 ) of grbs and being less affected by systematics .
simulations based on realistic extrapolations of ongoing ( and future ) grb experiments ( e.g. , , konus wind , glast ) show that : i ) the uncertainties on cosmological parameters can be significantly decreased ; ii ) future data will allow us to get clues on the `` dark energy '' evolution .
[ firstpage ] gamma rays : observations gamma rays : bursts cosmology : cosmological parameters . |
fmos ( the fiber multi - object spectrograph ) is a near - infrared ( nir ) fiber - fed spectrograph for the subaru telescope , which is capable of collecting nearly 400 spectra in a @xmath1 deg@xmath2 field - of - view ( @xcite ) available at the prime - focus of the telescope .
the 400 fibers are divided into two groups of 200 fibers and connected to the two spectrographs , irs1 ( infra - red spectrograph ) and irs2 , for spectroscopy .
a feature of fmos is the suppression of bright oh - airglow emission lines using oh mask mirrors , which reduce the oh - airglow emission lines by more than @xmath11 ( @xcite ) .
since these airglow lines are the largest noise source in the nir region , the oh suppression allows fmos to perform nir spectroscopic observations with a substantially reduced background level .
nir spectroscopy is a useful probe for a variety of topics in astronomy .
the detection of emission lines ( e.g. , h@xmath3@xmath12 , [ nii]@xmath13 , [ oiii]@xmath14 ) ) in galaxies provide us with information on diverse galaxy properties including the redshift , star formation rate , metallicity and ionisation state .
a cosmological redshift survey , fastsound , is ongoing , using fmos to measure @xmath15 redshifts of star - forming galaxies at @xmath16 from h@xmath3 lines over a total area of @xmath17 ( @xcite ) .
the main science goal is the measurement of redshift space distortion ( rsd ) for the first time in this redshift range , to measure the structure growth rate as a test of general relativity on cosmological scales ( see , e.g. , @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite for recent galaxy redshift surveys whose scientific targets include rsd ) .
a key for studies treating many emission line galaxies in a large data set is to automatically search and detect emission lines . in the case of fmos ,
residual oh airglow lines just outside the oh mask regions and cosmic rays often show spectral shapes similar to those of real emission lines .
if such spurious line detections are included into a statistical sample of emission lines , they will cause a systematic error in , e.g. , measurements of galaxy clustering power spectra .
therefore , an efficient automatic line detection software with minimized probability of false detection is highly desirable . in this paper
, we present the development of an emission line detection software for fmos spectroscopy , field ( fmos software for image - based emission line detection ) .
two - dimensional spectral images for each fiber are used in this algorithm in order to detect faint emission line features , while efficiently filtering out unwanted spurious detections .
we examine the performance of this routine by applying it to a part of the fastsound project data .
we test various parameters about shapes of emission lines in the 2d spectral images , to cut spurious objects efficiently .
the contamination fraction of spurious objects is estimated by inverted 2d spectral images , which are obtained by exchanging object frames and sky frames . in section [ sample ]
, we describe the data used in this work . in section [ method ] , we present the algorithm for detecting emission line candidates and describe key features that effectively reduce false detections caused by residual oh emission lines and cosmic rays .
the results are presented in section [ results ] and we summarize this work in section [ conclusion ] .
the targets for fmos spectroscopy were selected using photometric redshift and h@xmath3 flux estimates using @xmath18 data of canada - france - hawaii telescope legacy survey ( cfhtls ) wide ( @xcite ; @xcite ) .
observations are done with the normal - beam - switch ( nbs ) mode of fmos , in which the object frame is taken using nearly all @xmath0 fibers for targets , followed by an offset - sky frame with the same fiber configuration . in fastsound ,
the exposure time for one frame is @xmath19 s , and two frames ( i.e. , total exposure of @xmath20 min ) are taken both for the object and sky frames .
we use the high - resolution ( hr ) mode with spectral resolution of @xmath21 and the wavelength range of @xmath22@xmath23 m .
the detector size is 2k@xmath242k , and the pixel scale is @xmath25}$ ] .
the primary fmos data set used in this paper are eight fmos field - of - views ( fovs ) observed during march 2829 and april 12 of 2012 , as part of the fastsound project .
the fastsound field ids are w2_041 , 062 , and 063 from the cfhtls w2 field , and w3_162 , 163 , 164 , 177 , and 192 from w3 .
( see a forthcoming fastsound paper tonegawa , in prep . for details of the fastsound observations . )
fmos fibers were allocated to @xmath26 galaxies in the eight fmos fovs and the emission line detection rate is typically @xmath27 as reported in tonegawa et al .
the mean seeing is @xmath28 in the @xmath29 band .
these eight fovs are representative of fastsound data in normal observing conditions . in rare cases ( four out of @xmath30 fovs in the fastsound project ) ,
images taken by irs2 spectrograph just after sunset show unusual dark patterns on the edge of detector , likely because of the instability of detector resulting from overexposures by dome - flat lamp or twilight .
these patterns usually appear vertically in the left end ( @xmath22@xmath31 m ) of the image , but sometimes appear horizontally in the top or bottom end . though the fraction of such data is not large in the total fastsound data set , line detection must be affected in these regions . hence we also examine the performance of our software in these four fovs in these bad conditions .
these are w2_042 ( taken in 2013 feb . ) , w3_173 ( 2012 apr . ) , w3_206 and 221 ( 2012 jun . ) .
the fmos two - dimensional raw data are reduced with the standard data reduction pipeline , _ fibre - pac _ ( iwamuro et al , 2012 ) . in fmos images ,
the @xmath32-axis is used for the wavelength direction , and @xmath33 pixels are assigned in the @xmath34-axis direction for each fiber .
the profile in the @xmath34-axis is just the point spread function ( psf ) of light from a fiber .
first , sky subtraction is performed for each frame . in order to trace the time - variance of sky background
, the linear combination of two sky frames is used for the subtraction .
after sky subtraction , the bias difference between the four detector quadrants , cross talk , bad pixels , distortions , and residual oh lines are corrected .
the two images are then combined into the total exposure .
the wavelength calibration is applied by making a correlation between @xmath32-axis and physical wavelength using images of a th - ar lamp .
_ fibre - pac _ creates not only a science frame from object and sky images , but also square - noise and bad pixel maps for each fov ( figure 1 ) .
the square - noise map gives the square of noise level of each pixel , which is measured using @xmath35 pixels around each pixel along the @xmath34-axis with a @xmath36 clipping algorithm iterated ten times .
the bad pixel map indicates the quality of each pixel by a value between between @xmath37 ( bad ) and @xmath38 ( normal ) .
those pixels on the detector flagged as non - functioning by _ fibre - pac _ and temporally prominent pixels appearing in one exposure ( by cosmic ray events in most cases ) are rejected and replaced in the science frames , by an interpolated value from surrounding pixels ( @xcite ) .
these pixels are stored as a bad pixel map for each exposure of @xmath39 min .
our exposure time is @xmath20 min , and hence we have two bad pixel maps for each field - of - view .
when these two maps are combined , true defects of the detector are presented as the value @xmath37 on the bad pixel map because they appear in both of two bad pixel maps , while cosmic rays have a value of @xmath40 because they affect only one of the two science sub - frames .
the information from the square - noise and bad pixel maps are utilized in the line detection algorithm presented below . as well as the normal science frame
, we also create an inverted " science frame by exchanging object and sky frame in the _ fibre - pac _ procedures .
the line detection algorithm is also applied to this inverted science frame .
any candidates detected in this inverted frames will not be genuine features , because real emission lines are negative in the frame and absorption lines are usually below the detection limit of fmos for galaxies beyond @xmath41 .
therefore we can reduce the overall false detection rate by tuning the software to minimize the detection rate in the inverted frames .
( 80mm,80mm)figure1a.eps ( 80mm,80mm)figure1b.eps our goal is to automatically detect spectral features corresponding to emission lines from the fmos data with high reliability . to achieve this , we develop an emission line detection algorithm which is based on a convolution of the fmos 2-d spectral images weighted by a detection kernel that is similar to the typical emission line profile ( see , e.g. , @xcite for similar approaches ) .
note that we do not use the flux - calibrated 1-d spectra for emission line detection because they do not retain the 2-d shape of the psf , which is useful to filter out false detections .
the algorithm uses a flat - field image , square - noise map and bad pixel map as well as the science frame ( see [ sample ] ) . the science frame and square - noise map are both convolved with the kernel , producing an effective signal - to - noise ( @xmath6 ) ratio of a line centered at each pixel along the wavelength direction .
the flat - field image and bad pixel map are used to remove the oh mask regions or bad pixels from the calculation of @xmath6 , so that the false detection rate is minimized
. an important feature of the fmos spectra in the nir is the hardware suppression of the the oh airglow .
inclusion of the masked regions does not improve the line detection efficiency , and the noise level is often particularly high at the border of oh masks , which would have a negative effect for efficient line detection . to remove such regions , we utilize the dome - flat image .
the dome - flat image is divided by the detector - flat image to correct for different quantum efficiency between pixels ( @xcite ) .
after normalization , the flat - field image have values of @xmath42 at normal pixels , while @xmath43 at oh mask regions .
we then make a new square - noise map by dividing the original square - noise map by this flat image .
this operation increases the noise level inside and around oh mask regions by a large factor , making such regions ineffective in the line @xmath6 calculation ( see the next subsection ) .
this approach is better than , e.g. , simply removing oh mask regions by wavelength information , because the detailed performance of the oh mask depends on the temperature of fmos instrument that changes with time .
however , the effect of noisy regions around the border of oh masks can not be completely removed by this operation , and hence we further remove the pixels whose original noise level ( before dividing by the flat image ) is higher than the mean by more than @xmath44 , from the line @xmath6 calculation . in order to decrease the chance of detecting cosmic rays and detector defects , we use of the bad pixel map ( figure [ figure : fmosimage ] ) . since the bad pixel map has values of @xmath42 at good pixels , @xmath43 at defects , and @xmath45 at pixels hit by cosmic rays , we exclude the pixels having values lower than @xmath46 in the bad pixel map from the line @xmath6 calculations .
finally , the continuum component of galaxies may affect the detection efficiency of emission lines .
therefore the continuum component of each spectrum in the object frame was subtracted by applying fit1d task of iraf with a 5th - order chebyshev polynomial along the wavelength direction .
this procedure was adopted separately for each pixel along the vertical direction perpendicular to wavelength .
we define the detection kernel by a two - dimensional gaussian that imitates the typical shape of an emission line : @xmath47\ ] ] where @xmath48 and @xmath49 are the typical dispersion of emission lines along @xmath32-axis ( corresponding to wavelength ) and @xmath34-axis ( corresponding to the fiber aperture ) on the detector .
although @xmath48 and @xmath49 are adjustable , we fix @xmath50 [ pix ] as the quadratic sum of the spectral resolution of fmos ( @xmath51 fwhm at @xmath52 m in the hr mode ) and the typical velocity dispersion of galaxies ( @xmath53 fwhm ) , and @xmath54 [ pix ] as representative of the fmos fiber profile on the detector . because fmos is a fiber - fed spectrograph , @xmath49 is just an instrumetal spread of light from fibers without spatial information of target objects . for each pixel @xmath55 , we define the signal @xmath56 of a supposed line centered at the pixel as : @xmath57 which is an integration over pixels @xmath58 around @xmath55 confined within the region @xmath59 .
this region is an ellipse whose center is @xmath55 and the radii along the major and minor axis are @xmath60 [ pix ] and @xmath61 [ pix ] .
the parameter @xmath3 is the best - fit flux normalization of this kernel to the observed count @xmath62 in the pixels in @xmath59 : @xmath63 which can be derived by minimizing @xmath64 , where @xmath65 ^ 2}{n_i^2 } \ , \ ] ] where @xmath66 is @xmath67 noise at @xmath68-th pixel from the noise - square ( @xmath69 ) map . from the expression for @xmath3
, it can be understood that the signal @xmath56 ( equation [ s ] ) is a convolution of the count @xmath62 with the detection kernel , weighted by @xmath70 .
we then define the noise @xmath71 for the line ; since @xmath56 is a linear combination of @xmath62 whose noise is @xmath66 , its statistical error can be calculated as : @xmath72 then the signal - to - noise ratio @xmath6 is simply calculated from @xmath56 and @xmath71 . the @xmath6 is calculated along the wavelength direction , with @xmath73 fixed at the central pixel ( @xmath74 ) among the 9 pixels for one fiber .
when a certain pixel has a local peak of @xmath6 map , the pixel is likely to be the center of an emission line feature .
therefore , we select emission line candidates by extracting pixels with @xmath6 values higher than a threshold and locally greatest within the range of @xmath7530 pixels from @xmath76 along the wavelength direction . the latter condition is introduced to avoid multiple detections of the same line - like feature .
although the use of @xmath7530 pixels as the minimum separation is a simplistic approach , it suffices for fastsound , because the closest lines ( h@xmath3@xmath12 and [ nii]@xmath13 ) at @xmath16 are separeted by @xmath77 [ pix ] on the image .
more sophisticated methods to discriminate multiple lines , such as connected - pixel approach , will be examined in future work .
the selection threshold can be changed by the user : if one increases the threshold , the number of false detections decreases but the number of real lines would also be reduced .
the false detection rate and its dependence on the threshold are discussed in [ results ] .
the image shape of each candidate line includes important information that allows discrimination between real and false lines .
we measure the following shape parameters : \(1 ) center along @xmath34-axis : @xmath78 ( 2 ) dispersion along @xmath34-axis : @xmath79 ( 3 ) fraction of positive pixels : @xmath80 where @xmath81 is a step function ( 1 for @xmath82 and 0 otherwise ) .
+ ( 4 ) position angle of the major axis of an elliptical fit : @xmath83 ( 5 ) axis ratio of the elliptical fit : @xmath84 where @xmath85 and @xmath86 are semi - major and minor axes , respectively .
+ ( 6 ) signal - to - noise fluctuation per pixel : @xmath87 where angle brackets denote the mean over the region @xmath59 .
we applied our algorithm to an inverted " science frame , obtained by running the fmos image reduction pipeline with the object and sky frames exchanged , as well as the normal " science frame .
all the candidates in the inverted frames must be spurious , and their statistical nature should be the same as the spurious features in the normal frames , because the analysis procedures are exactly the same except for swapping the object / sky frames .
for example , residual oh emission lines in the science frame can be positive or negative at the same probability depending on the observing conditions , and cosmic rays fall randomly both on the object and sky frames .
note that an emission line galaxy can accidentally fall in a fiber during the sky exposures , but we did not take these cases into account , because the possibility should be small .
figure [ figure : param ] shows the distribution of the shape parameters defined above , as a function of @xmath6 , both for the normal and inverted frames .
it is found that in some regions ( especially for the data in the 4 fovs at the bad condition in the plots of @xmath88 , @xmath89 , @xmath90 , and @xmath91 plots ) the number of inverted - frame objects is relatively large compared with the normal - frame objects , and the numbers in normal / inverted frames are similar .
false lines are expected to be dominant over real lines in such regions , and hence we introduced an event cut as shown by the solid curves in the figure .
these conditions are expressed as : @xmath92 it should be noted that the region of many inverted - frame objects is also found in the plot of minor / major axis ratio .
we did not include a condition on this quantity , because we found that almost all of the false objects in this region are effectively removed by the other three conditions .
we also examined the spectral images of the rejected events by eye , and confirmed that false detections do indeed dominate .
these events mainly arise from unusual dark patterns appearing in the bad - condition four fovs , when the detector was in an unstable state .
examples of emission line candidates detected by the software are shown in figure [ figure : sample ] .
( 75mm,75mm)figure2a.eps ( 75mm,75mm)figure2b.eps ( 75mm,75mm)figure2c.eps ( 75mm,75mm)figure2d.eps ( 75mm,75mm)figure2e.eps ( 75mm,75mm)figure2f.eps ( 80mm,66mm)figure3a.eps ( 80mm,66mm)figure3e.eps ( 80mm,66mm)figure3b.eps ( 80mm,66mm)figure3f.eps ( 80mm,66mm)figure3c.eps ( 80mm,66mm)figure3g.eps ( 80mm,66mm)figure3d.eps ( 80mm,66mm)figure3h.eps
.detected line candidate statistics for the eight fovs under normal condition . [ cols="^,^,^,^",options="header " , ] [ table : linestat ] in table [ table : linestat ] , the number of emission line candidates in the eight fovs under normal condition is shown for three different thresholds of @xmath6 , both for the normal and inverted frames .
there is only one false - detection in the inverted frame above @xmath93 , indicating that the false detection rate should be less than 1% at @xmath8 .
we inspected the seven detections in the inverted frame at @xmath94 by eye , and found that five were close to oh mask region , one at the edge of the detector , and the other one due to a cluster of hot pixels that shows a collective instability because of high dark current .
although the line detection algorithm suppresses the contribution to @xmath6 from the oh mask region , strong residual oh emission often remains because of unstable oh airglow in the nir , resulting in a small fraction of false detections . since the events at the border of oh masks are the major source of the false detections , more strict cuts in such regions may further improve the fraction of real lines .
the cumulative @xmath6 distribution of emission line candidates are displayed in figure [ figure : snhist ] .
the lower limit to the number of real emission lines can be estimated from the difference between the normal and inverted frames , i.e. , red and blue lines .
( they are lower limits because the line detection completeness is not 100% by statistical fluctuation and oh mask effects .
see the next paragraph . )
one can see that the number of detections sharply increases with decreasing @xmath6 under @xmath95 , but this is mainly due to the increase of false detections in the inverted frame .
the wavelength of the detected lines are plotted against @xmath6 in figure [ figure : zhist ] .
again , we see a tendency for detections in the inverted images to be close to the oh mask regions , indicating that residuals from the oh airglow is one of the major reasons for false detections .
the completeness of detection , i.e. , the probability of successful detection for a given real emission line is also an important statistic to evaluate the performance of the software . to estimate this we ran a simulation by placing artificial objects and then applying our detection algorithm .
the completeness against input brightness ( @xmath6 ) is displayed in figure [ figure : completeness ] , for some different values of detection @xmath6 thresholds .
the completeness does not reach 100% but stays lower than 90% at large input @xmath6 , and this can be explained by the effect of oh masks that cover about 20% of the observed wavelength range .
indeed , we confirmed that the completeness becomes close to @xmath96 when we carried out the same simulation excluding the oh mask regions .
the completeness is about 40% for the input @xmath6 same as the adopted @xmath6 threshold , which is a reasonable result expected from statistical fluctuation and the oh mask effect .
it should be noted that we assumed the same gaussian psf as the kernel described in [ subsection : algorithm ] for the artificial objects distributed in the simulation .
though the kernel is set to be similar to typical line profiles , the completeness for real emission galaxies is expected to be lower than this .
a quantitative estimate of the completeness including this effect would be model dependent about the line profile distribution of a galaxy sample , which is beyond the scope of this work .
( 80mm,80mm)figure4a.eps ( 80mm,80mm)figure4b.eps ( 160mm,80mm)figure5.eps ( 80mm,80mm)figure6.eps
in this work we developed an automated emission line detection software for fmos on the subaru telescope , using the data from the fastsound project , targeting h@xmath3 emission line galaxies at @xmath16 .
emission line detection is based on @xmath6 obtained by a convolution between the two - dimensional science frame and a line - profile kernel .
a unique feature of the software for the fmos data having many oh airglow suppression masks is the amplification of noise level ( and consequent suppression in the @xmath6 calculation ) in the masked region by use of a flat - field image .
bad pixels on the detectors and pixels affected by cosmic - rays are removed from the @xmath6 calculation .
we also calculated six shape parameters for the detected lines , and these are used for further rejection of false detections .
this is particularly useful to remove those spurious sources caused by unusual dark patterns when the detectors are in unstable condition .
the efficiency and reliability of the line detection were examined by applying the method to inverted science frames obtained by exchanging the object and sky images in the reduction process .
the false - detection rate is @xmath97 , @xmath98 and @xmath99 for @xmath6 above @xmath100 , @xmath101 , and @xmath102 , respectively .
the emission line flux corresponding to @xmath103 is about @xmath104}$ ] in normal condition .
the software is open to the community , and currently available on request to the authors .
blake , c. , 2011 , , 415 , 2876 beutler , f. , 2012 , , 423 , 3430 de la torre , s. , guzzo , l. , peacock , j. a. , 2013 , , 557 , a54 guzzo , l. , 2008 , nature , 451 , 541 gilbank , d. g. , baldry , i. k. , balogh , m. l. , glazebrook , k. , & bower , r. g. 2010 , , 405 , 2594 gwyn , s. d. j 2012 , a&a , 143 , 38 hawkins , e. , 2003 , , 346 , 78 iwamuro , f. , motohara , k. , maihara , t. , hata , r. , & harashima , t. 2001 , , 53 , 355 iwamuro , f. , 2012 , , 64 , 59 kimura , m. , maihara , t. , iwamuro , f. , 2010 , , 62 , 1135 reid , b. a. , 2012 , , 426 , 2719 samushia , l , percival , w. j & raccanelli , a 2013 , , 420 , 2102 tonegawa , m. , totani , t. , akiyama , m. , 2014 , , 66 , 43 | we describe the development of automated emission line detection software for the fiber multi - object spectrograph ( fmos ) , which is a near - infrared spectrograph fed by @xmath0 fibers from the @xmath1 deg@xmath2 prime focus field of view of the subaru telescope .
the software , field ( fmos software for image - based emission line detection ) , is developed and tested mainly for the fastsound survey , which is targeting h@xmath3 emitting galaxies at @xmath4 to measure the redshift space distortion as a test of general relativity beyond @xmath5 . the basic algorithm is to calculate the line signal - to - noise ratio ( @xmath6 ) along the wavelength direction , given by a 2-d convolution of the spectral image and a detection kernel representing a typical emission line profile .
a unique feature of fmos is its use of oh airglow suppression masks , requiring the use of flat - field images to suppress noise around the mask regions .
bad pixels on the detectors and pixels affected by cosmic - rays are efficiently removed by using the information obtained from the fmos analysis pipeline .
we limit the range of acceptable line - shape parameters for the detected candidates to further improve the reliability of line detection .
the final performance of line detection is tested using a subset of the fastsound data ; the false detection rate of spurious objects is examined by using inverted frames obtained by exchanging object and sky frames .
the false detection rate is @xmath7% at @xmath8 , allowing an efficient and objective emission line search for fmos data at the line flux level of @xmath9[erg / cm@xmath2/s ] . |
almost a decade elapsed since the first detection of the anisotropy of the cosmic microwave background at large angular scales ( @xmath2 ) @xcite , @xcite . today
the cmb anisotropy ( cmba ) has been detected also at intermediate ( @xmath3 ) and small angular scales ( @xmath4 ) , so the cmba angular spectrum is now reasonably known down to the region of the first and second doppler peaks @xcite , @xcite , @xcite .
its shape gives information e.g. on the spectrum of the primordial cosmological perturbations or can be used to test the inflation theory but rises new questions to which cmba can not answers .
it is however possible to get answers looking at the cmb polarization ( cmbp ) .
for instance one can use cmbp to disentangle the effects of fundamental cosmological parameters like density of matter , density of dark energy etc .
this is among the goals of space and ground based experiments like @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite and is the main goal of sport a polarization dedicated asi / esa space mission on the international space station @xcite .
the relevance of the cmb polarization was remarked for the first time by m. rees @xcite . since his paper many models of the expected features of the cmbp
have been published ( see for instance @xcite , @xcite , @xcite ) .
they immediately stimulated the search for cmbp , but the first detection has been obtained only a few months ago ( @xcite and @xcite ) . because of its importance this discovery must be confirmed by new observations made with different systems and using different methods of extraction of the cmbp from the sky signal .
the detection of the cmbp is in fact extremely difficult because the signal is at least an order of magnitude below the cmba level .
moreover polarized foregrounds of galactic origin may cover the cmbp and/or mimic cmbp spots by their inhomogeneities .
the signal to noise ( cmbp @xmath5 polarized foreground ) ratio is therefore @xmath6 . in this paper
we discuss a method which improves this ratio and allows to disentangle cmbp and polarized foregrounds . in the microwave range the galactic foregrounds include : * synchrotron radiation ( strong polarization ) ,
* free - free emission ( null or negligible polarization ) , * dust radiation ( polarization possible ) beccause here we are interested in polarization , in the following we will neglect free - free emission.the effects of dust , if present , ( e.g. @xcite , @xcite ) , will be added to the synchrotron effects .
in fact , as it will appears in the following , what matters in our analysis are the statistical properties of the spatial distribution of the foregrounds and , by good fortune , the spatial distribution of the dust polarized emission is similar to that of the synchrotron emission . behind both types of radiation
there is in fact the same driving force , the galactic magnetic field which alignes dust grains and guides the radiating electrons ( @xcite and references therein ) . for the anisotropy
the separation of foregrounds and cmb was successfully solved by the dmr / cobe team when they discovered cmba @xcite .
the separation of cmbp and foregrounds is more demanding and recognizing true cmb spots among foreground inhomogneities severe .
approaches commonly used ( e.g. @xcite , @xcite , @xcite , @xcite , @xcite , + @xcite ) , are based on the differences between the frequency spectra of foregrounds and cmb , therefore require multifrequency observations . in this paper
we suggest a different method .
it takes advantage of the fact that , as we will show in the following , at small angular scales the values of the parameters used to describes the polarization of the diffuse radiation at a given frequency in different directions fluctuate .
we propose of analyzing the angular distribution of the polarized radiation on single frequency maps of the diffuse radiation and disentangling the main components , polarized synchrotron and cmbp , looking to their different statistical properties .
this method was proposed and briefly discussed in @xcite . here
we present a more complete analysis .
convenient quantities commonly used to describe the polarization status of radiation are the stokes parameters ( see for instance + @xcite , @xcite , + @xcite ) .
let s assume a monochromatic plane wave of intensity i and amplitude @xmath7 . in the _ observer plane _
, orthogonal to the direction of propagation of the electromagnetic wave , we can choose a pair of orthogonal axes @xmath8 and @xmath9 .
on that plane the amplitude vector of an unpolarized wave moves in a random way . on the contrary
it describes a figure , the _ polarization ellipse _
, when the wave is polarized . projecting the wave amplitude on @xmath8 and @xmath9 we get two orthogonal , linearly polarized , waves of intensity @xmath10 and @xmath11,(@xmath12 ) whose amplitudes are @xmath13 and @xmath14 respectively .
if the original wave of intensity @xmath15 is polarized , @xmath10 and @xmath11 are correlated : lets call @xmath16 and @xmath17 their correlation products . by definition
the stokes parameters are : @xmath18 , @xmath19 , @xmath20 , and @xmath21 .
@xmath22 and @xmath23 describe the linear polarization , @xmath24 the circular polarization and @xmath15 the total intensity .
the ratio @xmath25 gives the angle @xmath26 between the vector @xmath27 and the main axis of the polarization ellipse ( @xmath28 ) .
rotation by an angle @xmath29 of the @xmath30 coordinate system give a new coordinate system @xmath31 in which the stokes parameters become @xmath32 so when @xmath33 the axes of the polarization ellipse coincide with the reference axes @xmath34 and @xmath35 ( see fig.[f2 ] ) . to analyze the properties of the cmb polarization
it is sometimes convenient to use rotationally invariant quantities , like the radiation intensity @xmath15 and two combinations of @xmath23 and @xmath22 : @xmath36 and @xmath37 .
the intensity @xmath15 can be decomposed into usual ( scalar ) spherical harmonics @xmath38 .
@xmath39 the quantities @xmath40 can be decomposed into @xmath41 spin harmonics + @xcite , @xcite , @xcite + @xmath42 : @xmath43 the @xmath41 spin harmonics form a complete orthonormal system ( see , for instance , @xcite,@xcite , @xcite , @xcite ) and can be written @xcite , @xcite : @xmath44 where @xmath45 is a generalized jacobi polynomial , @xmath46 and : @xmath47 is a normalization factor .
the harmonics amplitudes @xmath48 correspond to the fourier spectrum of the angular decomposition of rotationally invariant combinations of stokes parameters . because spin @xmath41 spherical functions form a complete orthonormal system : @xmath49
we can write @xmath50 following @xcite and the very nice introduction made more recently by @xcite we now introduce the so called _ @xmath51 ( electric ) _ and _ @xmath52 ( magnetic ) modes _ of these harmonic quantities : @xmath53 they have different parities .
in fact when we transform the coordinate system @xmath54 into a new coordinate system @xmath55 , such that @xmath56 the e and b modes transform in a similar way : @xmath57 @xmath22 remains identical in both reference systems and @xmath23 changes sign .
it is important to remark that @xmath58 and @xmath59 are uncorrelated . in terms of @xmath22 and @xmath23
we can write : @xmath60 therefore : @xmath61 here @xmath62 and @xmath63 designate values of delta correlated 2d stochastic fields @xmath22 and @xmath23 .
omitting mathematical details , the correlation equations for @xmath22 and @xmath23 are : @xmath64 where @xmath65 is the dirac delta - function on the sphere .
( in the following we will sometimes omit indexes @xmath66 and @xmath67 ) .
synchrotron radiation results from the helical motion of extremely relativistic electrons around the field lines of the galactic magnetic field ( see , for instance @xcite , @xcite , @xcite ) .
the electron angular velocity @xmath68 is determined by the ratio between @xmath69 , the component of the magnetic field orthogonal to the particle velocity , and @xmath70 , the electron energy . as it moves around the magnetic field lines the electron radiates .
a)_single electron _
until the electron velocity is small ( @xmath71 ) we speak of cyclotron radiation : the electron behaves as a rigid dipole which rotates with gyrofrequency ( [ gyr1 ] ) in a plane orthogonal to the magnetic field direction and emits a single line . the radiation has a dumbell spatial distribution ( see fig.[f3 ] ) : @xmath72 is circularly polarized along the dumbell axis ( @xmath73 ) and linearly polarized in directions orthogonal to it ( @xmath74 ) .
when the electron velocity increases the radiation field changes until at @xmath75 , ( @xmath70 @xmath76 ) it assumes the peculiar charachters of synchrotron radiation : i)radiated power proportional to @xmath70 @xmath77 and @xmath78 , ii)continous frequency spectrum , peaked around : @xmath79 ( @xmath80 is the angle between the velocity vector @xmath81 and the wave vector @xmath82 , see fig.[f4 ] ) .
the peak is so narrow that in a given direction the emission is practically monochromatic , iii)radiation almost entirely emitted in a narrow cone ) of the cyclotron dumbbell beam , seen by a fast moving observer ( @xmath75 ) becomes a cone folded around the direction of movement ] of aperture ( see fig.[f5 ] ) @xmath83 around the forward direction of the electron motion . inside the cone
( @xmath84 ) the frequency is maximum and equal to @xmath85 in the opposite direction ( @xmath86 ) intensity and frequency are sharply reduced .
iv)radiation 100 @xmath87 linearly polarized at the surface of the cone.inside the cone the linear polarization is still dominant but a small fraction of circular polarization exists , ( @xmath88 and @xmath89 ) . outside the cone the very small fraction of radiation produced is elliptically polarized and becomes circularly polarized when seen along the @xmath69 direction .
so the stokes parameters depend on @xmath90 , the angle @xmath91 between @xmath90 and the line of sight , and the dimensionless frequency @xmath92 , where @xmath93 is the so called _ critical frequency _ , @xcite,@xcite , itself function of @xmath90 ( see eq.[gyr1 ] ) .
b)_cloud of monoenergetic electrons _ when the effects of many monoenergetic electrons with uniform distribution of pitch angles are combined , @xmath94 and @xmath23 are reinforced ( the stokes parameters are additive ) while @xmath24 is erased .
in fact @xmath95 where @xmath96 and @xmath97 are constants , @xmath98 and @xmath99 almost monochromatic functions of @xmath100 , and @xmath26 the angle between the projection of the magnetic vector on the observer plane and an axis on that plane ( the projection of the magnetic vector on the observer plane is the minor axis of the polarization ellipse ) . if inside the emitting cloud the magnetic field
varies , also @xmath26 varies , therefore @xmath101 and @xmath102 must be averaged along the line of sight across the cloud . in conclusion
the degree of polarization @xmath103 varies between a maximum value , ( uniform magnetic field ) and zero ( magnetic field randomly distributed ) .
c)_electrons with power law energy spectrum _ in the interstellar medium the radiating particles are the cosmic ray electrons whose energy spectrum : @xmath104 is a power law ( see for instance @xcite , @xcite ) and references therein ) with spectral index @xmath105 and space density proportional to @xmath106 . because the emission of a single electron is practically monochromatic
the resulting radiation spectrum is a power law : @xmath107 with intensity spectral index @xmath108 , temperature spectral index @xmath109 and @xmath110 a slow function of @xmath111 @xcite .
( if the magnetic field is not uniform we use @xmath112 instead of @xmath69 and a slightly different function @xmath113 ) .
the stokes parameters are products of the intensity @xmath114 , the degree of polarization @xmath103 and @xmath101 or @xmath102 , therefore we can write : @xmath115 where @xmath116 is a constant whose value depends on @xmath111 and the distribution of the magnetic field along the integration path . when the medium inside the synchrotron source , or in the medium where radiation propagates , is permeated by thermal electrons faraday rotation modifies the polarization charachteristics of the radiation .
in fact by faraday effect , when the radiation crosses a region of thickness @xmath117 permeated by magnetic fields and thermal electrons with density @xmath118 , the angle of polarization of the radiation rotates by an angle ( see for instance @xcite ) @xmath119 inside the source this brings depolarization , because radiation produced at different points along @xmath117 suffers different rotations .
additional depolarization inside the source comes about when the magnetic field along @xmath117 is not uniform because in eqs.([spec2])we have to use @xmath120 and @xmath121 instead of @xmath102 and @xmath101 .
so the degree of polarization at the source @xmath103 varies between 0 ( random magnetic field distribution ) and : @xmath122 ( uniform magnetic field and faraday rotation absent ) . outside the source the angle of polarization rotates by farday effect so the stokes parameter of the synchrotron radiation @xmath123 and @xmath124 measaured by an earth observer are different from the stokes parameters at the source . a regular trend of the magnetic field inside and outside the source is insufficient to guarantee a regular trend of the spatial distribution of @xmath123 and @xmath124 .
in fact as the line of sight moves among adjacent points on the sky @xmath118 , @xmath125 and @xmath117 fluctuate .
so also the angle of faraday rotation @xmath126 fluctuates by a quantity @xmath127 ^ 2 ~+~ [ < n_e > \delta(l)]^2\}^{1/2 }
~\simeq~cost~\lambda^2~l~\delta(n_e ) \end{array}\ ] ] proportional to @xmath117 .
because l is at least tens or hundred of pc @xmath128 can be very large .
besides this effect produced by faraday rotation , when we look in different directions through the interstellar medium @xmath123 and @xmath124 may change also because @xmath90 , @xmath129 , @xmath111 , @xmath130 and @xmath106 vary ( see for instance @xcite , @xcite ) .
is a nonzero mean , nonzero variance variable ; @xmath131 and @xmath132 are zero mean , nonzero variance variables ] we conclude that angle and degree of polarization of the synchrotron radiation fluctuates when the line of sight moves among adjacent regions on the sky , a conclusion supported by recent observations which show that the spatial distribution of the polarized component of the synchrotron radiation at intermediate and small angular scales is highly structured . on maps we see in fact lines ( _ channels _ ) along which @xmath103 goes to zero and sudden rotations of the plane of polarization when the line of sight goes from a side to the other of a channel @xcite .
we expect therefore that at small and intermediate angular scales @xmath123 and @xmath124 are stochastic functions of the direction of observations so we can write : @xmath133 exceptions to this behaviour can be expected when along the line of sight there are peculiar regions characterized by special field configurations and/or by large densities of thermal electrons , like in the well known loops and spurs or in hii regions , where systematic effects overcome random effects .
nature and origin of these features , clearly visible on brouw and spoelstra maps of the polarized component of the galactic background @xcite , are discussed for instance by @xcite .
usually these regions are close to the observer and have large angular dimensions , so their effects can be removed if one evaluates the angular power spectrum of the radiation distribution and limits his analysis to angular scales below few degrees ( multipole order @xmath134 ) .
being aware that this is a very important issue we tested its validity analyzing the distribution of the measured values of @xmath22 and @xmath23 on maps extracted from the effelsberg survey ( @xcite and @xcite ) of the diffuse radio emission .
we found ( see appendix 2 ) that up to angular scales of 5 degrees ( l close to 36 ) the difference @xmath135 is fully consistent with zero . because of eq .
( [ main ] ) if @xmath136 also @xmath137 , so the synchrotron radiation has both electric and magnetic modes : @xmath138 and ( see eqs .
( [ syn1 ] ) and ( [ main ] ) ) : @xmath139
in a homogeneous and isotropic universe only temperature and intensity @xmath140 change as the universe expands : both decrease adiabatically . because this is true for @xmath10 and @xmath11 separately , we do not expect anisotropy nor polarization therefore @xmath141 and @xmath142 are natural consequences . on the contrary , inhomogeneities and perturbations of matter density or of gravitational field , induce anisotropy and polarization of the cmb . at
the recombination epoch linear polarization appears as a by product of the thomson scattering of the cmb on the free electrons of the primordial plasma .
the polarizarion tensor it gives can be calculated solving the boltzman transfer equations of the radiation in a nonstationary plasma permeated by a variable and inhomogeneous gravitational field @xcite , @xcite , @xcite , @xcite .
the gravitational field is made of a background field , with homogeneous and isotropic frw metric , and an inhomogeneous and variable mix of waves : density fluctuations , velocity fluctuations , and gravitational waves . because of their transformation laws these waves are also said scalar , vector and tensor perturbations , respectively .
scalar ( density ) perturbations affect the gravitational field , the density of matter and its velocity distribution .
they were discovered studying the matter distribution in our universe on scales from @xmath143 mpc to @xmath144 mpc .
it is firmly believed they are the seeds of the large scale structure of the universe and are reflected by the large scale cmb anisotropy detected for the first time at the beginning of the 90s @xcite , @xcite .
their existence is predicted by the great majority of models of the early universe .
observation shows that the effects of these perturbations are small , so we can treat them as small variations @xmath145 ,
@xmath146 , @xmath147 of @xmath10 , @xmath11 and @xmath148 . introducing the auxiliary functions @xmath149 and @xmath130 : @xmath150 ( @xmath91 is the angle between the line of sight and the wave vector ) for plane waves the boltzman equations ( see for instance @xcite , @xcite , @xcite and reference therein ) become : @xmath151 where @xmath98 is the gravitational force which drives both anisotropy and polarization , @xmath152 is the thomson cross - section , @xmath118 is the density of free electrons , and @xmath153 the scale factor .
these equations give : @xmath154 where @xmath155 is the optical depth of the region where the phenomenon occurs .
rotating the coordinate system we can generate a new pair of stokes parameters ( @xmath156 ) : no matter which is the system of reference we choose these parameters satisfy the symmetry parity condition . because there is always a system in which @xmath157 and @xmath158 , we may conclude that , in a system dominated by the primordial density perturbations ( see , for instance @xcite ) magnetic modes of the cmb polarization vanish and only electric modes exist @xcite . therefore : @xmath159 where index @xmath160 stays for _ density perturbation_. vector perturbations , associated to rotational effects , perturb only velocity and gravitational field .
they are not predicted by the inflation theory and it is common believe that they do not contribute to the anisotropy and polarization of the cmb . gravitational waves ( tensor modes ) and gravitational lensing of large scale structures of the universe induce b modes in the distribution of cmbp(see for instance @xcite ) ) .
gravitational lensing gives a power spectrum at least an order of magnitude below the power spectrum of the e mode polarized signal produced by scalar perturbations , with a maximum at @xmath161 .
the power spectrum of the b mode polarization produced by gravity waves is maximun at @xmath162 , is definitely below the power spectrum produced by gravitational lensing for @xmath163 and only at very large angular scales ( @xmath164 ) it may be comparable to the scalar e modes . if one excludes very large angular scales we may conclude that cmbp is dominated by e - modes .
b - modes are just a contamination by b - modes at levels of @xmath165 or less .
the cmb radiation we receive is mixed with foregrounds of local origin .
when the anisotropy of the cmb was detected , to remove the foregrounds from maps of the diffuse radiation , data were reorganized in the following way : @xmath166 where @xmath167 , is a two dimension vector ( map ) which gives the total signal measured at different points @xmath168 on the sky , @xmath169 the noise vector , @xmath170 the matrix which combines the components @xmath171 of the signal . at each point
@xmath168 we can in fact write : @xmath172 where @xmath173 are weights , given by @xmath174 , @xmath175 is the synchrotron component , @xmath176 the free - free emission component , @xmath177 the dust contribution and so on . using just one map the signal components can not be disentangled .
if however one has maps of the same region of sky made at different frequencies it is possible to write a system of equations . provided the number of maps and equations is sufficient , the system can be solved and the components of @xmath178 separated , breaking the degeneracy .
we end up with a map of @xmath179 which can be used to estimate the cmb anisotropy .
when we look for polarization at each point on the sky we measure tensors instead of scalar quantities , therefore to disentangle the polarized components of the cmb we need a greater number of equations . here
we will concentrate on the separation of the two dominant components of the polarized diffuse radiation : galactic synchrotron ( plus dust ) foreground and cmbp . instead of observing the same region of sky at many frequencies , we suggest a different approach .
it takes advantage of the differences between the statistical properties of the two most important components of the polarized diffuse radiation ( cmb ( background ) and synchrotron ( foreground ) radiation ) and does not require multifrequency maps .
we define the estimator : @xmath181 where @xmath182 and @xmath183 ( here and in the following indexes @xmath184 or @xmath185 stay for _ synchrotron _ and cmb , respectively ) . because * @xmath51 and @xmath52 modes of synchrotron do not correlate each other neither correlate with the cmb modes + @xmath186 , + @xmath187 , * @xmath188 * @xmath189 where indexes @xmath160 and @xmath190 stay for @xmath191 ( or @xmath192 ) and @xmath193 perturbation , respectively .
@xmath180 gives an estimate of the e - mode excess in maps of the polarized diffuse radiation .
if tensor perturbations are negligible this excess is the cmbp signal .
if tensor perturbations are important the excess is a lower limit with a systematic difference from the true value which in the worst condition ( maximum contribution to cmbp from gravitational waves and gravitational lensing ) reaches a maximum value of @xmath165 .
let s now consider the angular power spectrum of @xmath180 . for multipole @xmath66
we can write : @xmath194 where @xmath195 and @xmath196 are random variables with gaussian distribution @xmath197 ( see eqs.([gauss1 ] ) and ( [ gauss2 ] ) in appendix a ) .
according the ergodic theorem ( in the limit of infinite maps , the average over 2d space is equivalent to the average over realisations ) the average value of @xmath198 is equal to the difference of the average values of @xmath195 and @xmath196 summed over @xmath67 . taking into account equation ( [ gauss2 ] ) we can therefore write : @xmath199 where : ) is an explicit form of the average of the stochastic variables @xmath200 and @xmath201 over a probability density @xmath197 , the short form being triangle brackets ] @xmath202 comparing eq.([coco ] ) with the ordinary definition of multipole coefficients : @xmath203 we can write : @xmath204 for synchrotron radiation @xmath205 should be zero , non zero for density perturbations , but on real maps it is always different from zero .
in fact a map is just a realization of a stochastic process and the amplitudes of @xmath195 and @xmath196 , averaged over 2d sphere , have uncertainties which add quadratically , so @xmath206 even in the case of synchrotron polarization .
this effect , very similar to the well known _ cosmic variance _ of anisotropy @xcite , @xcite ( the real universe is just a realisation of a stochastic process , therefore there will be always a difference between the realization we measure and the expectation value ) does not vanish if observations are repeated . the variance of @xmath198 is @xmath207 the quantities @xmath58 and @xmath59 , being sums over @xmath67 of @xmath208 stochastic values with gaussian distribution , have a @xmath209 distribution with @xmath210 degrees of freedom , so their variance is @xmath211 more explicitly @xmath212 when the synchrotron foreground is dominant ( @xmath213 ) @xmath214 in agreement with ( [ var1 ] ) . the synchrotron foreground is a sort of _ system noise _ which hampers the detection of the _ signal _ , the cmb polarization . at frequencies sufficiently high ( above @xmath215 ghz ,
see next section ) the noise is small compared to the signal therefore direct detection of cmbp is possible . at low frequencies on the contrary
the cmbp signal is buried in the noise created by synchrotron and dust emission . in this case to recognize the presence of the cmb polarization we can use our estimator @xmath180 . at angular scale @xmath66 , to be detectable the cmbp must satisfy the condition @xmath216 where @xmath217 are the coefficients of the multipole expansion of the @xmath51 modes and @xmath218 is the confindence level of the signal detectability . in a similar way we can write for our estimator : @xmath219 where @xmath220 if one neglects tensor perturbations ( @xmath221 ) the criterium for the cmbp detection becomes @xmath222 so by @xmath180 we get the cmbp level with an uncertainty @xmath223 which decreases as @xmath66 and the angular resolution increase
. tensor perturbations , if present , add to this uncertainty a systematic uncertainty @xmath224 .
the angular power spectra of the polarized component of the synchrotron radiation have been studied by @xcite , @xcite , @xcite , @xcite , @xcite and @xcite using the few partial maps of the polarized diffuse radiation available in literature .
it appears that the power spectra of the degree of polarization @xmath103 and of electric and magnetic modes @xmath51 and @xmath52 follow power laws of @xmath66 up to @xmath225 . for the degree of polarization
the spectral index is @xmath226 . for the @xmath51 and @xmath52 modes different authors get different values of the spectral index . according the authors of papers @xcite and @xcite ,
parkes data give : @xmath227 with @xmath228 and dependence of @xmath149 on the region of sky and the frequency . in paper @xcite , using effelsberg and parkes data , the authors get : @xmath229 with @xmath230 , @xmath231 and @xmath232 ( here we adjusted the original expression given in @xcite writing it in adimensional form )
. in paper @xcite the authors , using a completely different set of observational data ( @xcite and @xcite ) conclude that in the multipole range @xmath233 the spectral indexes of the @xmath51 and @xmath52 modes are @xmath234 and @xmath235 respectively , while for the polarized intensity the spectral index is @xmath236 . inside the multipole range @xmath237 the values of @xmath238 obtained by the three groups are marginally consistent and reasonably close to @xmath239 .
at lower value of @xmath66 definite differences exist between the results obtained by @xcite and by @xcite : these differences must be understood , but they are not important here because our statistical method can not be used for small values of @xmath66 when the number of samples becomes insufficient to carry on statistical analyses . for high values of @xmath66 , if one excludes extrapolations and models ( eg . @xcite ) , there are no data in literature , but this is not a limitation because for large values of @xmath66 the results of our method are practically independent from the spectral index .
all the above authors got their results analyzing low frequency data ( 1.4 , 2.4 and 2.7 ghz ) , therefore the extension of their spectra to tens of ghz , the region where cmb observations are usually made , depend on the accuracy of @xmath130 , the temperature spectral index of the galactic synchrotron radiation .
a common choice is @xmath240 but in literature there are values of @xmath130 ranging between @xmath241 and @xmath242 .
moreover @xmath130 depends on the frequency and the region of sky where measurements are made ( see @xcite , @xcite , @xcite , @xcite ) .
last but not least many of the values of @xmath130 in literature have been obtained measuring the total ( polarized plus unpolarized ) galactic emission . in absence of faraday effect @xmath243 and @xmath244 , the spectral indexes of the total , polarized and unpolarized components of the galactic emission , should be identical ( see eqs .
( [ spec1]),([spec2 ] ) ) .
however when faraday effect with its @xmath245 frequency dependence is present , we expect that , as frequency increases , the measured value of the degree of polarization ( se eq .
( [ degreep ] ) ) increases , up to @xmath246 therefore we should observe @xmath247 .
the expected differences are however well inside the error bars of the data in literature so at present we can neglect them and assume @xmath248 . instead of extrapolating low frequency results it would be better to look for direct observations of the galactic emission and its polarized component at higher frequencies . unfortunately above 5 ghz observations of the galactic synchrotron spectrum and its distribution are rare and incomplete . at 33 ghz observations by @xcite
give at some patches on the sky a galactic temperature of about @xmath249 from which follows that at the same frequency we can expect polarized foreground signals up to several @xmath250 . at 14.5 ghz observations made at ovro @xcite give synchrotron signals of 175 @xmath250 , equivalent to about 15 @xmath250 at 33 ghz of which up to 10 @xmath250 can be polarized .
in conclusion there are large uncertainties on the frequency above which observations of the cmbp are practically unaffected by the polarized component of the galactic diffuse emission .
to be on the safe side we can set it at 50 ghz ( see for instance @xcite , @xcite , @xcite , @xcite and @xcite ) . above 50 ghz cmbp
definitely overcomes the polarized synchrotron foreground .
below 50 ghz contamination by the galactic emission can be important but its evaluation , usually made by multifrequency observations , is dubious .
this conclusion is supported by figure [ f9 ] where we plotted , versus the multipole order @xmath66 , the power spectra of cmbp and galactic synchrotron at 43 ghz .
the cmb spectrum has been calculated by cmbfast @xcite assuming standard cosmological conditions ( cmb power spectrum normalized to the cobe data at low @xmath66 , @xmath251 , @xmath252 , @xmath253 , @xmath254 , @xmath255 km / sec / mpc , @xmath256 , @xmath257 , standard recombination ) .
the synchrotron spectrum has been calculated assuming for @xmath258 and @xmath259 the scaling law ( [ synebb ] ) with @xmath260 ( most pessimistic case ) , @xmath261 and @xmath262 respectively .
it appears that at 43 ghz the cmb power is comparable to the synchrotron power only at very small angular scales ( @xmath263 ) .
similar calculations at other frequencies confirm that only above @xmath0 ghz and at small angular scales ( large values of @xmath66 ) the cmbp power spectrum overcomes the synchrotron spectrum . below @xmath0 ghz direct detection of the cmbp is almost impossible even at small angular scale . to overcome this limit we can use our estimator . to see how it improves the cmbp detectability we calculated the angular power spectrum of @xmath264 , the estimator we expect when the diffuse radiation is dominated by synchrotron radiation . from eqs.([multipole ] ) , ( [ dl21 ] ) and ( [ synebb ] ) we get : @xmath265 a quantity which can be directly compared with the power spectrum of cmbp .
in fact eqs.([aed ] ) , ( [ d1 ] ) , ( [ equal ] ) and ( [ dl20 ] ) show that the power spectrum @xmath266 of the estimator evaluated when the sky is dominated by cmbp density perturbations coincides with the power spectrum of cmbp produced by density perturbations .
figure [ fd ] shows at 37 ghz the power spectra at 37 ghz of : i)@xmath264 , the estimator for a sky dominated by galactic synchrotron ( solid line , calculated using eq.([synebb ] ) ) , ii)@xmath266 , the estimator for a sky dominated by cmbp .
it coincides with the power spectrum of cmbp ( dotted line , calculated as in figure [ f9 ] with cmbfast using the same standard cosmological conditions ) , iii)the power spectrum of the polarized component of the galactic synchrotron radiation ( dashed line , calculated using eq.([synebb ] ) ) . as expected at 37 ghz
the cmbp power is well below the synchrotron power , therefore direct observations of cmbp are impossible ( the maximum value of the cmbp power is about 2.5 times below the synchrotron power at the same @xmath66 ) .
however above @xmath267 the power of the synchrotron estimator is definitely below the power of the cmbp estimator : at @xmath268 the ratio cmbp / d is maximum and close to @xmath269 .
this confirms that the use of @xmath180 allows to recognize the cmbp also at frequencies well below 50 ghz .
to further test the capability of our estimator we studied the separation of cmb and galactic synchrotron using measured instead of expected values of @xmath198.the measurements were simulated by random numbers , with gaussian distribution @xmath270 , zero mean and unity variance ( see eqs ( [ gauss1 ] ) - ( [ gauss2 ] ) ) .
two series of @xmath208 random numbers gave representations of @xmath271 and @xmath272 respectively and from them we obtained @xmath198 ( see eq.([dest ] ) ) .
then , to take into account that real data are collected with a finite angular resolution ( e.g. boomerang data come from regions whose angular extension is equivalent to @xmath273 @xcite , @xcite , @xcite ) we averaged the above values of @xmath198 on intervals @xmath274 : @xmath275 figure [ f6 ] , figure [ f7 ] and figure [ f8 ] show @xmath276 ( the sign of @xmath198 is arbitrary ) versus @xmath66 , for @xmath277 , @xmath278 and @xmath279 respectively : the very large fluctuations of the estimator are drastically reduced as soon as @xmath280 increases .
figure [ f10 ] , figure [ f11 ] and figure [ f12 ] are similar to figure [ fd ] . here , instead of the expectation value , we plot simulated measurements of @xmath264 at 37 ghz , for @xmath277 ( figure [ f10 ] ) , @xmath281 ( figure [ f11 ] ) and @xmath279 ( figure [ f12 ] ) , respectively .
once again the cmbp power spectrum comes from cmbfast assuming the same cosmological conditions we assumed above .
the synchrotron power spectrum is obtained from eq.([synebb ] ) with @xmath282 ( most pessimistic condition ) .
figure [ f13 ] shows at 17 ghz the same quantities we plotted in figure [ f12 ] . for a better appreciation of the differences among the three curves , on the vertical axis here we use a logarithmic scale .
the power spectrum of the estimator now almost touches the two highest peaks of the cmbp spectrum .
probably 17 ghz is the lowest frequency at which , in the most favorable conditions , one can use @xmath198 . in the most pessimistic case ( @xmath260 and @xmath283 )
the corresponding frequency is 25 ghz .
we analyzed the impact of the real world experimental conditions on our method for disentangling cmbp and foregrounds in maps of the polarized diffuse radiation .
a polarimeter is a two channel system which ( by hardware and/or software methods ) splits the sky signal in two polarized components , send them to separate channels where they are amplified and then looks for correlations between the two components .
the system outputs are proportional to a pair of stokes parameters ( e.g. @xmath23 and @xmath22 ) , or to the electric and magnetic modes @xmath58 and @xmath59 or to a combination of them , e.g. our estimator @xmath180 .
unfortunately as the signals propagate through the system , noise is added to them so at the system output the signal is mixed to noise .
moreover if the channels are asymmetric or there are cross talks between them , the outputs contain additional signals which simulate spurious polarization , usually an offset of the system output from the level one should expect with an ideal system when polarization is absent ( for a discussion see for instance @xcite ) .
so in the real world the signal we are looking for , @xmath284 , is accompanied by uncorrelated noise and offset .
what one measures is @xmath285 where @xmath286 , @xmath287 is the cross talk coefficient , @xmath288 the noise signal and @xmath289 the noise standard deviation . in commercial systems
@xmath290 while in dedicated cmbp experiments values of @xmath291 have been obtained @xcite . by accurate choice of the system components we can therefore minimize the offset and
further reduce it by phase modulation techniques ( see e.g. @xcite ) . at this point
we have noisy maps we can use to evaluate @xmath180 and its power spectrum instead of @xmath23 and @xmath22 or @xmath58 and @xmath59 .
as shown by the above simulation the signal / noise ratio for @xmath180 is definitely better than the signal / noise ratio for the stokes parameters or @xmath58 or @xmath292 and brings it to values similar to the ones we find when studying weak radiosources or , in the worst situation , the cmb anisotropy .
from here we can therefore go on and extract @xmath180 from the remaining noise using the well known methods of time integration , commonly used in radioastronomy .
observations of the cmb polarization are hampered by the presence of a foreground , the polarized component of the galactic synchrotron radiation . only above @xmath293
the cosmic signal definitely overcomes the galactic synchrotron and direct measurements of the cmbp are possible . between @xmath294 ghz and @xmath215 ghz
background and foreground are comparable .
below @xmath295 ghz the polarized sky is dominated by the galactic signal .
so when measurements are made at ground observatories where atmospheric absorption prevent observations above @xmath296 ghz , all programs for cmbp measurements must include observing and analysis strategies for disentangling the galactic synchrotron signals from the cmbp signals .
a common approach is fitting models of the intensity , frequency dependence and spatial distribution of the cosmic and galactic signals to multifrequency maps of the polarized diffuse radiation .
or , when observation are made at one frequency , looking for additional data in literature , but here the accuracy of the available data is insufficient to get firm evaluations of cmbp . in this paper
we presented a different approach which at small angular scales ( @xmath297 ( @xmath298 ) ) and down to frequencies as low as @xmath299 ( @xmath300 in the most favorable conditions ) allows to extract the cmbp signal from single frequency maps of the polarized diffuse radiation .
it takes advantage of the different statistical properties of the spatial distributions of cmbp and polarized galactic synchrotron . by our estimator @xmath180 , which evaluates the difference between e- and b - modes , we get the polarized component of cmb with a maximum systematic (
underestimate ) uncertainty of @xmath165 .
this uncertainty is set by the contamination by the tensor perturbations which add b - modes to a cmbp map dominated by the e - modes generated by scalar(density ) perturbations .
improving our knowledge of the tensor perturbations we will reduce the above uncertainty and improve the accuracy of our method of measuring cmbp .
the accuracy we can get with our method is the maximun one can obtain at ground observatories with today ( 2nd generation ) systems for measuring the cmb polarization .
these 2nd generation experiments are just arrived on the verge of detecting the polarized signals produced by density perturbations ( see for instance @xcite and @xcite).direct observations of the signals associated to tensor perturbations requires new , 3rd generation , intrinsically able to reject the foreground signals which , at present , are still in preparation ( see for instance @xcite ) .
we are indebted to s.cortiglioni , e.caretti , e.vinjakin , j.kaplan and j. delabrouille for helpful discussion .
mvs acknowledges the osservatorio astronomico di capodimonte , inaf , for hospitality during preparation of this paper .
here and overall in paper we suppose that @xmath301 are complex random variables which satisfy the probability distribution law : @xmath302 with variance @xmath303 and @xmath304 .
they have all the propertiers of gaussian variables ( below we omitt indexes @xmath51 and @xmath52 in first and second equations ) : @xmath305 @xmath306 setting @xmath307 with current index @xmath51 or @xmath52 , it immediately follows : @xmath308
in the our paper we assert that for the galactic synchrotron emission the measured values of the stokes parameters @xmath22 and @xmath23 behave as stochastic variables and random fields with gaussian distribution .
this statement certainly holds at high galactic latitudes . at low galactic latitudes , where we observe large scale galactic structures , regular magnetic fields @xcite and quasi - periodic structures with typical sizes of about 250 pc , ( the amplitudes of regular and irregular components are approximately equal )
this assumption has to be checked . to do it we analyzed the distribution of the measured values of @xmath22 and @xmath23 on regions of different extension extracted from the effelsberg maps of the polarized diffuse radiation @xcite .
the six fields were chosen within @xmath309 from the galactic plane , at galactic longitudes between @xmath310 and @xmath311 .
they had dimensions of @xmath312 , @xmath313 , @xmath314 , @xmath315 , @xmath316 , @xmath317 , respectively .
characteristics of the observed distributions of @xmath22 and @xmath23 measured at low galactic latitudes in areas of various extensions ( the larger is the number of pixels the closer is the histogramm to a gaussian shape ) [ cols="<,^,^,^,^,^ " , ] for each region we examined the shapes of the distributions , and calculated average value and the variance of the measured values of @xmath22 and @xmath23 .
the distribution of @xmath22 in a @xmath318 area is shown in figure [ fa1 ] .
very similar results are obtained for @xmath23 and for smaller regions .
when the region is very small , ( figure [ fa2 ] is for an @xmath312 area ) , the distribution becomes broad , but its shape is still gaussian .
results of analyses carried on on all the areas are presented in table 1 : for sufficiently large samples both @xmath22 and @xmath23 are compatible with zero . where @xmath288 is the number of pixels in the field , @xmath320 is the measured variance of @xmath22 and @xmath23 and @xmath321 is our estimator .
the quantity ( [ appeq1 ] ) is plotted in figure [ fa3 ] versus the number @xmath288 of pixels . in the same figure
we plot also : i)the calculated values of the differences @xmath322 in each field and ii)the calculated values of @xmath323 | the polarization of the cosmic microwave background ( cmb)is a powerful observational tool at hand for modern cosmology .
it allows to break the degeneracy of fundamental cosmological parameters one can not obtain using only anisotropy data and provides new insight into conditions existing in the very early universe .
many experiments are now in progress whose aim is detecting anisotropy and polarization of the cmb .
measurements of the cmb polarization are however hampered by the presence of polarized foregrounds , above all the synchrotron emission of our galaxy , whose importance increases as frequency decreases and dominates the polarized diffuse radiation at frequencies below @xmath0 ghz . in the past the separation of cmb and synchrotron
was made combining observations of the same area of sky at different frequencies . in this paper
we show that the statistical properties of the polarized components of the synchrotron and dust foregrounds are different from the statistical properties of the polarized component of the cmb , therefore one can build a statistical estimator which allows to extract the polarized component of the cmb from single frequency data also when the polarized cmb signal is just a fraction of the total polarized signal .
our estimator improves the signal / noise ratio for the polarized component of the cmb and reduces from @xmath150 ghz to @xmath120 ghz the frequency above which the polarized component of the cmb can be extracted from single frequency maps of the diffuse radiation .
, , polarimetry , mathematical procedures , radio and microwave , observational cosmology : 95.75.hi , 95.75.pq , 95.85.bh , 90.80.es |
systems consisting of a single atom coupled to a high - finesse optical cavity are of fundamental importance to quantum optics and quantum information science .
such cavity qed systems have been experimentally implemented using neutral atoms @xcite and ions @xcite , and have been the subject of numerous theoretical studies @xcite . in particular , such systems play a key role in proposals for scalable quantum computation @xcite and distributed quantum networks @xcite .
an important requirement for many of these proposals is the ability to coherently control the internal and motional degrees of freedom of the trapped atom , and raman transitions provide the means for meeting this requirement .
raman transitions are powerful tools that have diverse applications in atomic physics , including spectroscopy @xcite , precision measurement @xcite , and coherent state manipulation @xcite , and have been used to coherently control the motional degrees of freedom of trapped ions @xcite and of neutral atoms trapped in optical lattices @xcite . until recently
, however , raman transitions had not been incorporated into cavity qed .
the practical challenge of implementing a free - space raman scheme in cavity qed lies in the presence of the cavity itself , which offers limited optical access to the atom within , especially in the strong coupling regime @xcite .
as first proposed in @xcite , these optical access limitations can be circumvented by implementing a cavity - based raman scheme , in which the raman coupling is generated by standing - wave fields inside the cavity . in this paper
, we provide the theoretical background behind two such cavity - based schemes for driving raman transitions between the hyperfine ground state manifolds of a single atom optically trapped within a high - finesse cavity .
both schemes have been recently implemented and validated experimentally , and have been used to extend the trapping lifetime of a near - resonantly driven atom @xcite , to cool the axial motion of an atom to the quantum ground state of its trapping potential @xcite , and to optically pump an atom into a specific zeeman state @xcite .
the paper is organized as follows . in section [ sec : trapping ] , we describe how a single atom is trapped inside the cavity by means of an optical dipole trap . in section [ sec : raman - three - level ] , we present the two schemes for driving raman transitions in the trapped atom . we treat the atom using a three - level model , and show that the internal and motional degrees of freedom of the atom can be described using an effective hamiltonian that has the same form for both schemes . in section [ sec : quantize - axial ] ,
we quantize the axial motion of the atom , and show how the raman couplings allow one to drive transitions that change the vibrational quantum number for motion along the cavity axis .
we present both analytic results , which apply to atoms that are sufficiently cold , and numerical results , which apply to atoms of arbitrary temperature . in section [ sec : raman - cesium ] , we describe how the raman couplings are modified when we take into account the multiplicity of levels in a physically realistic cesium atom .
we show that the raman schemes drive transitions between individual zeeman states in the two hyperfine ground state manifolds , and we derive the rabi frequencies for these zeeman transitions . finally , in section [ sec : cooling ] , we show how raman transitions can be used to cool the axial motion of the atom to the quantum ground state , and we present computer simulations of the cooling process .
the optical cavity we will be considering consists of two symmetric mirrors separated by a distance @xmath0 ( details of the cavity are given in appendix [ sec : cavity - mode - structure ] ) .
an atom is trapped inside the cavity by means of a far off - resonance trap ( fort ) , which is created by driving one of the cavity modes with light that is red - detuned from a dipole transition in the atom .
the red - detuned light forms a standing wave inside the cavity , and the coupling of the atom to the light causes it to be attracted to the points of maximum intensity in the standing wave . the operation of the fort can be understood by considering a simple two - level model in which the atom has a single ground state @xmath1 and a single excited state @xmath2 .
let @xmath3 denote the frequency of the @xmath4 transition , and let @xmath5 denote the frequency of the fort light .
we will assume that the fort resonantly drives mode @xmath6 of the cavity , so @xmath7 , where @xmath8 is the free spectral range . the hamiltonian for the system is @xmath9 where @xmath10 here @xmath11 is the position of the atom , @xmath12 is a dimensionless quantity that characterizes the shape of the cavity mode ( see appendix [ sec : cavity - mode - structure ] ) , and @xmath13 is the rabi frequency of the light at a point of maximum intensity . we can simplify this hamiltonian by making the rotating wave approximation and then performing a unitary transformation to eliminate the time dependence : @xmath14 where @xmath15 is the detuning of the fort from the atom ( note that because the fort is red - detuned , @xmath16 ) .
in the limit that the fort is far - detuned ( @xmath17 ) , we can adiabatically eliminate the excited state ( see @xcite ) to obtain an effective hamiltonian for the ground state : @xmath18 where @xmath19 describes a trapping potential with depth @xmath20 .
for the experiments described in @xcite , the power in the fort beam is set such that @xmath21 . to describe the shape of the potential , it is convenient to use a cylindrical coordinate system centered on the cavity axis : we will let @xmath22 and @xmath23 denote the axial and radial coordinates of the atom , where the cavity mirrors are located at @xmath24 and @xmath25 .
using equation ( [ eqn : mode - shape - general ] ) to substitute for the mode shape @xmath12 in this coordinate system , we find that @xmath26 where @xmath27 is the wavenumber for the fort mode @xmath6 .
the minima of the potential occur at the points @xmath28 , @xmath29 , where @xmath30 is defined such that @xmath31 since @xmath32 , we find that @xmath33 ; thus , there are @xmath6 distinct fort wells in which an atom can be trapped .
let us assume that the atom is trapped in fort well @xmath34 .
it is convenient to define a coordinate @xmath35 that gives the axial displacement of the atom from the potential minimum of this well .
we can then express the trapping potential as @xmath36 near the bottom of the well the potential can be approximated as harmonic , @xmath37 where @xmath38 and @xmath39 , the radial and axial vibrational frequencies , are given by @xmath40 the corresponding periods @xmath41 and @xmath42 characterize the timescales for radial and axial motion . for the experiments described in @xcite , the vibrational frequencies are @xmath43 and @xmath44 , so the timescale for radial motion is much longer than the timescale for axial motion .
we will be interested in describing the evolution of system over timescales that are short compared to the timescale for radial motion , but not necessarily short compared to the timescale for axial motion .
for such timescales we can view the atom as being radially stationary , and take @xmath23 to be a constant parameter that enters into the potential for the axial motion .
thus , dropping a constant term , we can express the potential for the axial motion as @xmath45 where @xmath46 is the axial trap depth at radial coordinate @xmath23 .
to show how a raman coupling can be generated in a trapped atom , let us now consider a three - level model in which the atom has two ground states @xmath47 and @xmath48 , which correspond to the ground state hyperfine manifolds of a multi - level atom , and a single excited state @xmath2 .
the excited state has energy @xmath3 , and the ground states @xmath47 and @xmath48 have energies @xmath49 and @xmath50 , where @xmath51 is the ground state hyperfine splitting . the hamiltonian for the atom is @xmath52 where @xmath53 we can generate a raman coupling between the two ground states by driving one of the cavity modes with a pair of beams that are tuned into raman resonance with the ground state hyperfine splitting of the atom .
let us denote the optical frequencies of these beams by @xmath54 , where @xmath55 is the average frequency and @xmath56 is the frequency difference ( see figure [ fig : raman - config - levels ] ) .
also , let us define a parameter @xmath57 that gives the raman detuning of the beams ; we will assume that the beams are tuned close to raman resonance , so @xmath58 .
the beams generate standing - wave fields inside the cavity , and the coupling of the atom to these fields is described by the hamiltonian @xmath59 where @xmath60 and @xmath61 is an atomic lowering operator . here @xmath11 is the position of the atom , @xmath62 is a dimensionless quantity that characterizes the shape of the driven mode ( see appendix [ sec : cavity - mode - structure ] ) , and @xmath63 are the rabi frequencies of the fields at a point of maximum intensity . for simplicity , we have assumed that the @xmath64 and @xmath65 transitions couple to the light fields with equal strength .
the total hamiltonian for the system is @xmath66 .
we can simplify this hamiltonian by making the rotating wave approximation and then performing a unitary transformation : @xmath67 where @xmath68 and @xmath69 describes the overall detuning of the optical fields from the excited state .
we will assume that the fields are far - detuned from the atom ( @xmath70 ) , so we can further simplify the hamiltonian by adiabatically eliminating the excited state to obtain an effective hamiltonian for the ground states : @xmath71 where @xmath72 and @xmath73 we will assume that the optical fields are weak enough that @xmath74 , so the first term of @xmath75 dominates and to a good approximation the eigenstates of @xmath75 are @xmath76 and @xmath77 .
the second term of @xmath75 describes a state - independent level shift , which is analogous to the fort potential we derived in the previous section .
the third term of @xmath75 gives a state - dependent correction to the level shift , which is of order @xmath78 and may therefore be neglected .
the fourth term of @xmath75 describes a modulation of the state - independent level shift at frequency @xmath56 . because we have assumed that the system is tuned near to raman resonance , @xmath56 is of the same order as the hyperfine splitting @xmath51 . for cesium ,
the atom used in the experiments described in @xcite , the hyperfine splitting is @xmath79 , which is much larger than the harmonic frequencies @xmath38 and @xmath39 that characterize the timescales for atomic motion .
thus , over the motional timescales the fourth term of @xmath75 averages to zero and may also be neglected . after making these approximations ,
we are left with @xmath80 we can further simplify the effective hamiltonian by making the rotating wave approximation and then performing a unitary transformation to eliminate the time - dependence : @xmath81 this hamiltonian describes an effective two - level atom with ground state @xmath47 and excited state @xmath48 , which is driven by a classical field with rabi frequency @xmath82 and detuning @xmath83 .
we will now use this effective hamiltonian to describe two schemes for driving raman transitions in a trapped atom . and @xmath48 , the excited state @xmath2 , and the pair of optical fields at frequencies @xmath84 and @xmath85 .
] in the first scheme , which we will call the fort - raman configuration , the fort itself forms one leg of a raman pair . to form the other leg of the pair we add a much weaker beam , which we will call the raman beam , that drives the same cavity mode as the fort but is detuned from the cavity resonance by @xmath56 .
this configuration was first proposed in @xcite , and formed the basis of the optical pumping scheme described in @xcite and of the cooling scheme used in @xcite to extend the lifetime of a trapped atom .
let us denote the optical frequencies of the fort and raman beams by @xmath5 and @xmath86 , and their maximum rabi frequencies inside the cavity by @xmath13 and @xmath87 .
we can then apply the results of section [ ssec : raman - three - level - effective - hamiltonian ] to obtain an effective hamiltonian @xmath75 that describes the fort - raman pair .
we will assume that the raman beam is blue - detuned from the fort ( @xmath88 ) , so we take the @xmath89 field to be the raman field and the @xmath90 field to be the fort field ; if the raman beam is red - detuned from the fort ( @xmath91 ) then we reverse these identifications .
as was discussed in section [ sec : trapping ] , we treat the atom as being radially stationary , and consider only the axial motion .
thus , the total hamiltonian for the atom is given by @xmath92 here @xmath93 is the momentum of the atom in the axial direction , and @xmath75 is given by equation ( [ eqn : effective - hamiltonian ] ) with @xmath94 and @xmath95 , where @xmath96 and @xmath97 are the maximum level shifts due to the fort and raman fields individually , and @xmath98 is the overall detuning of the fort and raman beams from the atom .
it is convenient to express @xmath99 as @xmath100 , where @xmath101 describes the axial motion of the atom , and @xmath102 describes the internal state of the atom .
typically the powers of the fort and raman beams are such that @xmath103 and @xmath21 , and for these values @xmath104 .
thus , we can neglect the level shift of the raman beams and approximate @xmath105 as @xmath106 where @xmath107 is the fort trapping potential . as in section
[ sec : trapping ] , we will assume that the atom is trapped in fort well @xmath34 , and define a coordinate @xmath35 that gives the axial displacement of the atom from the well minimum . note that @xmath108 where we have substituted for the fort mode shape @xmath12 using equation ( [ eqn : mode - shape - general ] ) and for @xmath109 using equation ( [ eqn : fort - well - locations ] ) .
thus , we can express the internal hamiltonian as @xmath110 where @xmath111 is the effective rabi frequency for an atom at radial coordinate @xmath23 .
for the second configuration , which we will call the raman - raman configuration , the fort drives mode @xmath6 of the cavity at frequency @xmath5 , and a pair of raman beams drives mode @xmath112 of the cavity at frequency @xmath113 .
this configuration was used to perform the ground - state cooling described in @xcite .
we will assume that the raman beams have equal powers and are tuned symmetrically about the cavity resonance , so the frequencies of these beams can be expressed as @xmath114 .
let @xmath13 denote the maximum rabi frequency for the fort beam , and let @xmath87 denote the maximum rabi frequency for one of the raman beams . using the results of section [ ssec : raman - three - level - effective - hamiltonian ]
, we can describe the coupling of the atom to the pair of raman fields in terms of an effective hamiltonian @xmath75 . as was discussed in section [ sec : trapping ] , we treat the atom as being radially stationary , and consider only the axial motion .
thus , the total hamiltonian for the system is @xmath115 the first term describes the kinetic energy of the atom due to axial motion , and the second term describes the fort potential @xmath107 , where @xmath96 is the fort depth .
the third term is given by equation ( [ eqn : effective - hamiltonian ] ) with @xmath116 , @xmath117 , and @xmath118 , where @xmath119 is the overall detuning of the raman pair from the atom .
it is convenient to express @xmath99 as @xmath100 , where @xmath120 describes the motion of the atom , and @xmath121 describes the internal state of the atom .
note that because the fort and raman beams drive different cavity modes , the registration of the standing waves corresponding to the two beams depends on the axial position of the atom , and therefore on the particular fort well in which the atom is trapped .
this is in contrast to the fort - raman configuration , for which the fort and raman standing waves are always perfectly registered .
one consequence of the well - dependence of the registration is that the level shift @xmath122 due to the raman pair distorts different fort wells in different ways .
we calculate this effect in appendix [ sec : potential - distortion ] , but for now we note that for the typical parameters @xmath103 , @xmath21 the ratio of the level shifts due to the fort and raman beams is @xmath123 .
thus @xmath124 , so we can neglect the @xmath122 term and approximate @xmath105 as @xmath125 as in section [ sec : trapping ] , we will assume that the atom is trapped in fort well @xmath34 , and define a coordinate @xmath35 that gives the axial displacement of the atom from the well minimum .
note that @xmath126 where @xmath127 is the phase difference between the fort and raman beams at the bottom of fort well @xmath34 .
here we have substituted for the fort mode shape @xmath12 using equation ( [ eqn : mode - shape - general ] ) and for @xmath109 using equation ( [ eqn : fort - well - locations ] ) .
we will assume that the fort and raman beams drive nearby modes of the cavity , so @xmath128 .
in this limit @xmath129 and @xmath130 , so we can approximate equation ( [ eqn : raman - intensity - profile ] ) as @xmath131 thus , we can express the internal hamiltonian as @xmath132 where @xmath111 is the effective rabi frequency for an atom at radial coordinate @xmath23 .
we have shown that that for both the fort - raman and the raman - raman configurations the hamiltonian has the form @xmath100 , where @xmath133 for the fort - raman configuration @xmath134 , and for the raman - raman configuration @xmath127 for an atom trapped in fort well @xmath34 .
the hamiltonian @xmath105 describes the motion of the atom in the axial potential and is independent of the atomic state , and the hamiltonian @xmath135 describes a raman coupling between the ground states that depends on the axial position of the atom . for either the fort - raman or the raman - raman configuration , the raman coupling described by @xmath135 can be turned off by turning off one of the beams in the raman pair ( note that for fort - raman configuration the fort beam must always be on in order to maintain the trapping potential , so for this configuration one must turn off the raman beam ) .
alternatively , one can keep the beams in the raman pair on at all times and turn off the raman coupling by tuning the pair out of raman resonance ( @xmath136 ) .
as we have seen , there is a small level shift due to the raman beams , which we neglected when writing down the above expression for @xmath105 , and this detuning - based method has the advantage that these level shifts are always present regardless of whether the raman coupling is on or off . with the first method
, these level shifts cause a slight change in the trapping potential whenever the raman coupling is turned on or off , which could potentially heat the atom or cause other problems .
for many applications , such as raman sideband cooling , it is necessary to quantize the axial motion ; that is , to treat the axial position @xmath137 and momentum @xmath93 as quantum operators .
we first show how this is achieved for cold atoms , and then discuss some numerical results that apply to atoms of arbitrary temperature . for cold atoms
the axial trapping potential is nearly harmonic , where the harmonic frequency @xmath138 is given by @xmath139 we can quantize the axial motion by introducing phonon creation and annihilation operators operators @xmath140 and @xmath48 , which are related to @xmath137 and @xmath93 by @xmath141 from these relations , we find that @xmath142 where @xmath143 , the lamb - dicke parameter , is given by @xmath144 .
note that because @xmath138 depends on the radial coordinate @xmath23 , the lamb - dicke parameter also depends on @xmath23 .
if the atoms are sufficiently cold , we can obtain a reasonable approximation to @xmath145 by expanding in @xmath143 and retaining terms only up to second order ; from equations ( [ eqn : h - ext ] ) , ( [ eqn : h - int ] ) , and ( [ eqn : kf - x ] ) , we find that @xmath146 it is convenient to form a basis of states @xmath147 for the system by taking tensor products of the internal states @xmath76 and @xmath77 with the motional fock states @xmath148 . to order @xmath149 these product states are eigenstates of @xmath105 , where pairs of states @xmath150 , @xmath151 with the same vibrational quantum number @xmath152 are degenerate and have energy @xmath153 the raman coupling described by @xmath135 drives transitions between the product states . by taking matrix elements of the raman coupling , we find that state @xmath150 is coupled to states @xmath151 , @xmath154 , and @xmath155 , where the rabi frequencies for these transitions are given by @xmath156 note that @xmath157 transitions are suppressed relative to @xmath158 transitions by @xmath159 , and @xmath160 transitions are suppressed relative to @xmath158 transitions by @xmath161 . to resonantly drive the @xmath162 transition we set @xmath163 , and to resonantly drive the @xmath164 and @xmath165 transitions we set @xmath166 and @xmath167 , where @xmath168 and @xmath169 .
for a harmonic trap @xmath170 and the frequencies of these transitions are independent of @xmath152 , but because the fort is shallower than its harmonic approximation , @xmath171 decreases with increasing @xmath152 . recall that for the fort - raman configuration @xmath172 , whereas for the raman - raman configuration the value of @xmath173 depends on the fort well in which the atom is trapped .
thus , in the fort - raman configuration the rabi frequencies are the same for all the fort wells , whereas in the raman - raman configuration the rabi frequencies vary from well to well .
also , note that in the fort - raman configuration the @xmath157 transitions are always forbidden .
this follows from symmetry considerations : since the trapping potential is symmetric under @xmath174 the motional eigenstates are also parity eigenstates , and since the raman coupling is symmetric under @xmath174 it can not couple an even parity state to an odd parity state .
the harmonic approximation we described in the previous section only applies to fock states @xmath175 for which @xmath176 . for higher - lying energy levels
, we can calculate the rabi frequencies and detunings for the various motional transitions by numerically solving the time - independent schrdinger equation for @xmath105 .
this provides us with a set of motional eigenstates @xmath177 and eigenvalues @xmath178 . using the motional eigenstates
, we can take matrix elements of the raman coupling described by @xmath135 to calculate the rabi frequencies for different motional transitions : @xmath179 note that @xmath135 can also drive higher - order @xmath152-changing transitions , but the matrix elements for these transitions are quite small and we will not consider them here . from the energy eigenvalues , we can determine the detunings for the @xmath164 and @xmath165 transitions : @xmath180 the numerically - determined rabi frequencies and detunings are shown in figure [ fig : motional - parameters-1 ] for @xmath134 and @xmath181 , and in figure [ fig : motional - parameters-2 ] for @xmath182 .
note that for @xmath134 and @xmath181 the @xmath157 transitions are forbidden , so we only plot the rabi frequencies and detunings for the @xmath158 and @xmath160 transitions , and for @xmath182 the @xmath160 transitions are forbidden , so we only plot the rabi frequencies and detunings for the @xmath158 and @xmath157 transitions . for these graphs
the lamb - dicke parameter is taken to be @xmath183 , which is the value relevant for the experiments described in @xcite .
: green curve is @xmath184 , blue curve is @xmath185 , red curve is @xmath186 .
, title="fig : " ] : green curve is @xmath184 , blue curve is @xmath185 , red curve is @xmath186 .
, title="fig : " ] : green curve is @xmath184 , blue curve is @xmath187 , red curve is @xmath188 . ]
level diagram for cesium . ]
so far we have discussed the fort - raman and raman - raman schemes in the context of a simple three - level model . in this section ,
we show how these schemes are modified when we take into account the multiplicity of levels in a physically realistic alkali atom , using cesium as an example . a level diagram for cesium
is shown in figure [ fig : cesium - spectrum ] ; the levels relevant to our considerations include the ground state hyperfine manifolds @xmath189 and @xmath190 , which correspond to ground states @xmath47 and @xmath48 of the three - level model , and the excited state manifolds @xmath191 and @xmath192 , which correspond to the excited state @xmath2 of the three - level model .
the hamiltonian for a free cesium atom is @xmath193 where the sum is taken over all the states @xmath2 in the @xmath191 and @xmath192 excited state manifolds , and where @xmath194 and @xmath195 are projection operators onto the @xmath196 and @xmath197 ground state manifolds .
the quantity @xmath198 is the hyperfine splitting between the @xmath196 and @xmath197 ground state manifolds , and @xmath3 is the energy of excited state @xmath2 , where the zero of energy is taken to be halfway between the two ground state manifolds . as in section
[ ssec : raman - three - level - effective - hamiltonian ] , we want to derive the raman coupling that results when the atom is trapped within an optical cavity and one of the cavity modes is driven with a pair of beams that generate standing - wave fields inside the cavity .
the coupling of the atom to the standing - wave fields is described by a hamiltonian @xmath199 that has the same form as equation ( [ eqn : raman - pair - three - level ] ) , but with @xmath200 given by @xmath201 here @xmath202 are the maximum intensities of the two fields , and @xmath203 and @xmath204 are the spontaneous decay rate and saturation intensity for the @xmath191 excited state manifold .
we can express @xmath205 as @xmath206 where @xmath207 is the wavelength of the @xmath208 transition .
the quantities @xmath209 are the polarizations of the two fields , which we will take to be linear and mutually orthogonal .
thus , the vectors @xmath210 form an orthonormal frame , where @xmath211 is a unit vector that lies along the cavity axis .
the vector @xmath212 is an atomic lowering operator , and is defined by @xmath213 here @xmath214 is the clebsch - gordan coefficient that connects ground state @xmath215 to excited state @xmath216 via polarization @xmath217 , @xmath218 is a orthonormal basis of polarization vectors , and @xmath219 is a weighting factor for transitions between the @xmath220 and @xmath221 hyperfine manifolds , where the quantity in brackets is a @xmath222-symbol . following the procedure we used in section [ ssec : raman - three - level - effective - hamiltonian ] , we can adiabatically eliminate the excited states to obtain an effective hamiltonian for the ground states . because the derivation closely parallels the derivation given in section [ ssec : raman - three - level - effective - hamiltonian ]
, we will omit the intermediate steps and simply quote the result : @xmath223 where @xmath224 here @xmath225 is the overall detuning of the raman pair from excited state @xmath2 . in the limit that @xmath226 is much larger than the excited state hyperfine splittings , one can show that @xcite @xmath227,\end{aligned}\ ] ] where @xmath228 and @xmath229 are the overall detunings of the raman pair from the cesium @xmath230 and @xmath231 lines .
it is convenient to express these detunings as @xmath232 where @xmath233 are dimensionless parameters . from equations ( [ eqn : v - eff - sum ] ) , ( [ eqn : omega - eff - sum ] ) , and ( [ eqn : sum - over - excited - states ] ) , we find that @xmath234 where @xmath235 and @xmath236 is an atomic lowering operator that couples zeeman states in @xmath197 to zeeman states in @xmath196 .
if we collect these results , make the rotating wave approximation , and perform a unitary transformation to eliminate the time - dependence , we can express the effective hamiltonian as @xmath237 it is instructive to compare to the effective hamiltonian for the three - level model given in equation ( [ eqn : effective - hamiltonian ] ) with the effective hamiltonian for the full cesium atom given in equation ( [ eqn : effective - hamiltonian - cesium ] ) .
the two hamiltonians have similar forms , only the operator @xmath238 that coupled ground states @xmath47 and @xmath48 has been replaced by the operator @xmath239 that couples zeeman states within the ground state manifolds @xmath196 and @xmath197 .
in addition , we now have equations ( [ eqn : v - eff - cesium ] ) and ( [ eqn : omega - eff - cesium ] ) that allow us to calculate the parameters @xmath122 and @xmath240 in terms of the intensities of the standing - wave fields .
following the same reasoning that was used in sections [ sec : raman - raman - three - level ] and [ sec : fort - raman - three - level ] , we can use the effective hamiltonian given in equation ( [ eqn : effective - hamiltonian - cesium ] ) to write down the total hamiltonian @xmath99 for the fort - raman and raman - raman configurations . in both cases the total hamiltonian has the form @xmath100 , where @xmath241 here @xmath242 and @xmath243 are the axial trap depth and the effective rabi frequency at radial coordinate @xmath23 , and the parameters @xmath244 and @xmath240 are calculated for the fort - raman and raman - raman configurations in the following sections [ sec : fort - raman - cesium ] and [ sec : raman - raman - cesium ] . as was discussed in section [ sec : fort - raman - three - level ] , in the fort - raman configuration the fort forms one leg of the raman pair , and a weak raman beam is added to form the second leg .
the fort resonantly drives mode @xmath6 of the cavity , and the raman beam drives the same mode as the fort but is detuned from the cavity resonance by @xmath245 .
we can obtain expressions for the fort depth @xmath244 and the effective rabi frequency @xmath240 by using equations ( [ eqn : v - eff - cesium ] ) and ( [ eqn : omega - eff - cesium ] ) , which relate these quantities to the maximum intensities of the fort and raman beams inside the cavity , together with equation ( [ eqn : intracavity - intensity ] ) from appendix [ sec : cavity - mode - structure ] , which relates these maximum intensities to the optical powers of the fort and raman beams at the input of the cavity ( note that because the raman beam is detuned from the cavity resonance , its coupling into the cavity is suppressed ) .
we find that @xmath246 here @xmath247 and @xmath248 are the powers of the fort and raman beams at the input of the cavity , @xmath249 is a reference power that is set by the cavity geometry and is defined in equation ( [ eqn : power - c ] ) of appendix [ sec : cavity - mode - structure ] , @xmath250 is the total energy decay rate for the fort mode @xmath6 , and @xmath251 and @xmath252 , the detuning parameters at the fort wavelength @xmath253 , are given by equation ( [ eqn : detuning - parameters ] ) .
it is interesting to note that for fixed powers in the fort and raman beams , the effective rabi frequency @xmath240 monotonically increases as the cavity decay rate @xmath250 is reduced . in deriving the expression for @xmath244 given in equation ( [ eqn : fort - depth ] )
, we assumed that the detuning of the fort from the cesium @xmath230 and @xmath231 lines was the same for the @xmath196 and @xmath197 ground state hyperfine manifolds .
this is a reasonable approximation , because these detunings are much larger than the hyperfine splitting @xmath51 .
however , because the detuning of the @xmath196 manifold is slightly larger than the detuning of the @xmath197 manifold , the fort potential is slightly weaker for @xmath196
. thus , the fort squeezes the two manifolds together , causing a small reduction in the effective hyperfine splitting .
this effect , which is calculated in appendix [ sec : differential - stark ] , gives a slight position - dependence to the effective raman detuning , but this can be neglected for many applications .
as was discussed in section [ sec : raman - raman - three - level ] , in the raman - raman configuration the fort resonantly drives mode @xmath6 of the cavity , and pair of raman beams drives mode @xmath112 of the cavity .
we will assume that the two raman beams have equal powers @xmath248 and are tuned symmetrically about the cavity resonance .
the fort depth @xmath244 is given by equation ( [ eqn : fort - depth ] ) , and we can obtain an expression for the rabi frequency @xmath240 by using equation ( [ eqn : omega - eff - cesium ] ) , which relates the rabi frequency to the maximum intensities of the raman beams inside the cavity , together with equation ( [ eqn : intracavity - intensity ] ) from appendix [ sec : cavity - mode - structure ] , which relates these maximum intensities to the optical powers of the raman beams at the input of the cavity ( note that because the raman beams are detuned from the cavity resonance , their coupling into the cavity is suppressed ) .
we find that @xmath254 here @xmath249 is a reference power that is set by the cavity geometry and is defined in equation ( [ eqn : power - c ] ) of appendix [ sec : cavity - mode - structure ] , @xmath255 is the total energy decay rate for the raman mode @xmath112 , and @xmath256 and @xmath257 , the detuning parameters at the raman wavelength @xmath258 , are given by equation ( [ eqn : detuning - parameters ] ) .
note that for fixed powers in the raman beams there is an optimal cavity decay rate @xmath259 that maximizes the effective rabi frequency @xmath240 . the operator @xmath239 that appears in @xmath135 couples individual zeeman transitions between the @xmath196 and @xmath197 ground state hyperfine manifolds . in this section
, we calculate the matrix elements for these transitions .
let us introduce an arbitrary coordinate system @xmath260 and define a set of zeeman states @xmath261 relative to this coordinate system .
we can express the unit vector @xmath211 that lies along the cavity axis as @xmath262 where @xmath263 is the angle between the cavity axis @xmath211 and the quantization axis @xmath264 .
note that @xmath265 where @xmath266 are angular momentum raising and lowering operators .
thus , the state @xmath267 couples to states @xmath268 and @xmath269 , and the matrix elements corresponding to these transitions are @xmath270 where we have used that the matrix elements of @xmath271 are given by @xcite @xmath272 note that if the quantization axis is aligned along the cavity axis then @xmath273 transitions are forbidden , and if the quantization axis is transverse to the cavity axis then @xmath274 transitions are forbidden .
we have shown that the raman coupling can drive transitions that raise or lower the axial vibrational quantum number @xmath152 . in this section ,
we show how one can exploit these @xmath152-changing transitions to cool the axial motion to the vibrational ground state .
we will first show how the cooling works using the three - level model , and then discuss cooling for a physically realistic cesium atom .
one way to cool the atom is to alternate coherent raman pulses tuned to an @xmath152-lowering transition with incoherent repumping pulses . to see how this works ,
let us assume that we start out with the atom in state @xmath150 .
we can lower the vibrational quantum number by driving the atom with a coherent raman pulse tuned to the @xmath275 transition , which transfers some of the population from @xmath150 to @xmath276 .
the atom can then be repumped to ground state @xmath47 by driving the @xmath65 transition with near - resonant light . the repumping light drives the atom to the excited state , from which it spontaneously decays to either ground state @xmath48 , where it continues to interact with the repumping light , or to ground state @xmath47 , where it is dark to the light .
if the atom is sufficiently cold to begin with , then the repumping process is unlikely to change the atom s vibrational state , because the matrix elements for @xmath152-changing transitions are suppressed relative to the matrix elements for @xmath152-preserving transitions by at least @xmath277 , where @xmath278 .
thus , the net effect of the raman and repumping pulses is to move some of the population from state @xmath150 to state @xmath279 . by iterating the pulse sequence ,
the atom can be cooled to a state that has a mean vibrational quantum number @xmath280 that is close to zero .
the same type of scheme can be used to cool a multi - level alkali atom . for a cesium atom ,
the @xmath196 ground state manifold plays the role of state @xmath47 and the @xmath197 ground state manifold plays the role of state @xmath48 : we start with the atom in a random zeeman state in @xmath196 , drive the atom with a coherent raman pulse tuned to the @xmath281 transition , and then repump the atom to @xmath196 .
it is easiest to understand the effects of these pulses if we choose the quantization axis to lie along the cavity axis , so that only @xmath274 transitions are allowed and the raman coupling drives transitions between pairs of states @xmath282 .
if the ambient magnetic fields are nulled , then these zeeman transitions are all degenerate , so the transition frequency of the @xmath281 transition is independent of @xmath283 .
thus , the coherent raman pulse is effective at lowering the vibrational quantum number regardless of which zeeman state in @xmath196 the atom started in : each zeeman pair behaves equivalently , except for a slight @xmath283-dependence in the rabi frequency that is given by equation ( [ eqn : zeeman - zero ] ) .
the zeeman state of the atom is then scrambled during the repumping phase , so at the beginning of the next cooling cycle the atom starts out in a potentially new zeeman state . the amount of time it takes to cool the atom is determined by the amount of time it takes to repump the atom , which is set by the spontaneous decay rate of the excited state , and by the amount of time it takes to perform the coherent raman pulse , which is set by the rabi frequency @xmath284 .
the cooling rate can be increased by increasing the rabi frequency , but as we increase the rabi frequency we begin to off - resonantly drive the @xmath162 transition , and this sets an upper limit to the rabi frequency that can be used .
off - resonant driving of the @xmath285 transition becomes important when @xmath286 , so the upper limit to the rabi frequency is given by @xmath287 .
there is also a lower limit to the value of @xmath280 that can be achieved with this cooling scheme , which is set by two different factors .
first , when we resonantly drive the @xmath288 transition with the coherent raman pulse , we can also off - resonantly drive the @xmath289 transition .
this mechanism gives a lower limit of @xmath290 .
we can reduce this limit by reducing the rabi frequency , but since the rabi frequency determines the cooling rate , this also slows down the cooling .
in addition , there are problems with using small rabi frequencies that are due to the anharmonicity of the fort , which will be discussed later .
ideally , one would gradually reduce the rabi frequency as the atom cools , so as to balance the conflicting demands for a high cooling rate and a low value of @xmath280 . a second factor that limits @xmath280
is the fact that when the atom is repumped it will not always remain in the same vibrational state it started out in , since the lamb - dicke suppression of @xmath152-changing transitions is not perfect .
this mechanism gives a lower limit of @xmath291 .
the cooling scheme described above can be modified in several ways .
first , rather than alternating raman pulses with repumping pulses , it is also possible to continuously drive the atom with both raman and repumping light , and this is the method that was used in @xcite and @xcite .
second , the cooling scheme we described relies on @xmath292 transitions , but it is also possible to cool the atom using @xmath293 transitions . indeed , for the fort - raman configuration the @xmath292 transitions are forbidden , so the atom can only be cooled using @xmath293 transitions .
cooling via @xmath293 transitions tends to be slower than cooling via @xmath292 transitions , since the condition @xmath286 gives an upper limit on the rabi frequency of @xmath294 . also , for @xmath293 transitions
both the state @xmath295 and the state @xmath296 decouple from the raman pulse , so the state to which the system cools depends on the initial state : if we start in a state @xmath150 with @xmath152 even then the system cools to @xmath295 , and if we start in state @xmath150 with @xmath152 odd then the system cools to @xmath296 .
note that because the fort is anharmonic , the resonant frequency of the @xmath292 and @xmath293 transitions depends on the value of @xmath152 .
this means that if we keep the raman detuning @xmath83 set at at a fixed value throughout the cooling process , then the detuning of the raman pulse from the atom will change as the atom cools .
we can estimate the importance of this effect by considering @xmath292 and @xmath293 transitions as separate cases .
first we will consider @xmath292 transitions .
let us assume that we set the raman detuning to @xmath297 , so the detuning of the raman pulse from the @xmath275 transition is @xmath298 as we have shown , the maximum rabi frequency that can be used is @xmath299 , and for this maximum value the ratio of the detuning to the rabi frequency is @xmath300 , which is small for cold atoms .
thus , for @xmath292 transitions the dependence of the detuning on @xmath152 is a small effect ; we can simply set the raman detuning to @xmath301 , and as long as the atoms start out reasonably cold the cooling will always be efficient . now consider @xmath293 transitions .
we will assume that the raman detuning is set to @xmath302 , so the detuning of the raman pulse from the @xmath303 transition is @xmath304 as we have shown , the maximum rabi frequency that can be used is @xmath294 , and for this maximum value the ratio of the detuning to the rabi frequency is @xmath305 .
thus , for @xmath293 transitions the dependence of the detuning on @xmath152 is a significant effect . to compensate for this problem , one could slowly decrease the raman detuning @xmath83 during the cooling process to ensure that the raman pulse remains in resonance as the atom cools .
although we have focused on cooling the axial motion of the atom , it is possible to implement the axial cooling schemes in such a way that they cool the atom s radial motion as well .
this is accomplished by configuring the repumping light so that it provides polarization gradient cooling @xcite in the plane transverse to the cavity axis .
specifically , the repumping light is tuned blue of the @xmath306 transition , and is delivered to the atom via two pairs of counter - propagating beams .
the two pairs of beams are perpendicular to one another and to the cavity axis , and therefore provide cooling in both transverse directions .
one can characterize the effectiveness of the cooling schemes described in the previous section by using raman spectroscopy to measure the temperature of the atom . in what follows
, we will assume that @xmath292 transitions are used to cool the atom , although the same methods can also be applied to cooling via @xmath293 transitions . to measure a raman spectrum ,
we cool the atom , pump it into ground state @xmath47 , and then drive it with a coherent raman pulse with detuning @xmath83 .
we then check if the atom was transfered to @xmath48 . by iterating this sequence
one can measure the probability that the atom is transfered from @xmath47 to @xmath48 by the raman pulse , and by repeating this measurement for raman pulses of different detunings one can map out a raman spectrum . for an atom in vibrational state @xmath152 , the raman spectrum will exhibit a peak at @xmath163 , which corresponds to @xmath285 transitions , and peaks at @xmath307 , which correspond to @xmath164 transitions .
we will refer to the peak at @xmath163 as the carrier , and the peaks at @xmath308 and @xmath309 as the red and blue sidebands . because of the fort anharmonicity @xmath171 depends on @xmath152 , but we will assume that the atoms are cold enough that this effect can be neglected and simply take @xmath310 . one way to determine the axial temperature of the atom is to measure the ratio of the red to the blue sideband ; this is the same technique as was used in @xcite to determine the temperature of a trapped ion . for a thermal distribution ,
the probability that the atom has axial vibrational quantum number @xmath152 is given by @xmath311 where @xmath312 is the mean vibrational quantum number .
if we start with the atom in state @xmath47 and resonantly drive the blue sideband with a raman pulse of duration @xmath313 , the probability that the atom is transfered to state @xmath48 is given by @xmath314 if we start in state @xmath47 and resonantly drive the red sideband , the probability that the atom is transfered to state @xmath48 is given by @xmath315 note that @xmath316 so the ratio of the transfer probabilities for the red and blue sidebands is @xmath317 an alternative way to quantify the cooling is to measure the population in the vibrational ground state .
this can be accomplished by pumping the atom to state @xmath47 and applying a raman pulse whose detuning is adiabatically swept across the red sideband . if the atom started in the vibrational ground state @xmath318 then it will remain in state @xmath47 , and if the atom started in a vibrational state @xmath319 then the raman pulse will adiabatically transfer it to state @xmath48 . thus , the population in the vibrational ground state is given by the probability that the atom remains in state @xmath47 after the adiabatic sweep has been completed .
the advantage of this method is that it does not rely on assuming that the atoms are thermally distributed .
it is also possible to use raman spectroscopy to say something about the radial temperature : since the axial frequency @xmath320 depends on the radial coordinate @xmath23 , the width of the sidebands depends on the radial temperature .
the probability that the atom has axial frequency @xmath321 is given by @xmath322 where @xmath323 is the potential for radial motion , and @xmath324 if the radial temperature is small compared to the trap depth ( @xmath325 ) , then we can make a harmonic approximation and perform the integral analytically : @xmath326 thus , if the blue sideband has width @xmath327 , one can put an upper limit on the radial temperature of @xmath328 .
the cooling schemes discussed in section [ ssec : cooling - schemes ] can be simulated on a computer .
we will take the hamiltonian for the system to be @xmath329 where @xmath135 and @xmath105 are given by equations ( [ eqn : h - int ] ) and ( [ eqn : h - ext ] ) , and where @xmath330 describes the coupling of the atom to repumping light . here
@xmath331 is the rabi frequency of the repumping light and @xmath332 is the detuning of the light from the @xmath65 transition . as was discussed in section [ ssec : cooling - schemes ] , in order to radially cool the atom we use light that is blue - detuned from the @xmath333 transition to repump the atom . to model this in the simulation
, we will assume that the excited state decays to ground state @xmath47 at rate @xmath334 and to ground state @xmath48 at rate @xmath335 , where @xmath203 is the spontaneous decay rate for the @xmath191 manifold of cesium , and the prefactors @xmath336 and @xmath337 , the branching ratios for spontaneous decay on the @xmath338 and @xmath339 transitions , are given by equation ( [ eqn : weighting - factor ] ) . also , we will take @xmath340 to be the detuning that optimizes the polarization - gradient cooling .
we can write down a master equation for the system , which describes both the coherent evolution due to @xmath99 and the incoherent evolution due to the spontaneous decay from the excited state .
given an initial state , we can numerically integrate the master equation to obtain the state of the system at later times .
in figure [ fig : cooling-1]a , we use this method to simulate cooling in the raman - raman configuration : we start the system in state @xmath341 and plot @xmath280 as a function of time .
we assume that the atom is trapped in a fort well with @xmath182 , and use @xmath292 transitions to perform the cooling . for this simulation ,
the cooling parameters are @xmath342 , @xmath343 , @xmath344 . in figure
[ fig : cooling-1]b , we simulate cooling in the fort - raman configuration . in the fort - raman configuration @xmath134 for all the fort wells and the @xmath292 transitions
are forbidden , so we use @xmath293 transitions to perform the cooling .
as was previously discussed , this means that the asymptotic state to which the system cools depends on the initial state .
two curves are shown in the graph : for one , we start the system in state @xmath345 ; for the other , we start the system in state @xmath346 . for these simulations ,
the cooling parameters are @xmath342 , @xmath343 , @xmath347 . versus time @xmath313 , ( a ) cooling via @xmath292 transitions starting from state @xmath348 ,
( b ) cooling via @xmath293 transitions starting from states @xmath345 and @xmath346 .
, title="fig : " ] versus time @xmath313 , ( a ) cooling via @xmath292 transitions starting from state @xmath348 , ( b ) cooling via @xmath293 transitions starting from states @xmath345 and @xmath346 . , title="fig : " ] in addition to simulating the time - evolution of the system , we can calculate the asymptotic value of @xmath280 by solving the master equation for the steady - state density matrix .
this can be used to study the dependence of the asymptotic value of @xmath280 on the various cooling parameters . in figure
[ fig : cooling-2 ] , we consider cooling in the raman - raman configuration for atoms with @xmath349 , and plot the asymptotic value of @xmath280 as a function of @xmath350 , @xmath331 , and @xmath83 .
the parameters that are not being varied are set to the same values used for the cooling simulation shown in figure [ fig : cooling-1]a .
these graphs show that the cooling scheme is quite robust and works efficiently over a broad range of parameters . versus @xmath350 , ( b ) asymptotic @xmath280 versus @xmath331 , ( c ) asymptotic @xmath280 versus @xmath83 .
, title="fig : " ] versus @xmath350 , ( b ) asymptotic @xmath280 versus @xmath331 , ( c ) asymptotic @xmath280 versus @xmath83 .
, title="fig : " ] versus @xmath350 , ( b ) asymptotic @xmath280 versus @xmath331 , ( c ) asymptotic @xmath280 versus @xmath83 .
, title="fig : " ]
we have described two schemes for driving raman transitions in an atom trapped within a high - finesse optical cavity .
these schemes can be used to control both the internal and motional degrees of freedom of the atom , and provide powerful tools for studying cavity qed ; as an example , we have shown in detail how the raman schemes can be used to cool the atom to the quantum ground state of the trapping potential .
although the two schemes are similar in many respects , they do have some important differences .
the fort - raman scheme has the advantage that the raman coupling is independent of the fort well in which the atom is trapped , and is thus better suited for manipulating the internal state of the atom . on the other hand
, the raman - raman scheme has the advantage that the @xmath351 transitions are allowed for most fort wells , and is thus better suited for cooling .
the ability to coherently control the atom is a key requirement for many cavity qed protocols , and these raman schemes should open up new possibilities for experiments in cavity qed .
the author would like to thank a. boca , r. miller , and t. e. northup for helpful suggestions .
this research is supported by the national science foundation , the army research office , and the disruptive technology office of the department of national intelligence .
here we describe the mode structure of the optical cavity .
the cavity we will be considering consists of two symmetric mirrors of radius @xmath352 that are separated by a distance @xmath0 . it is convenient to define a cylindrical coordinate system that is centered on the cavity axis : we will denote the distance from the cavity axis by @xmath23 and the displacement along the cavity axis by @xmath22 , where the mirrors are located at @xmath24 and @xmath25 .
the cavity supports a set of discrete modes with resonant frequencies at integer multiples of the free spectral range @xmath8 , where for each frequency there are two degenerate modes corresponding to the two polarization states transverse to the cavity axis .
consider one of the polarization modes with mode order @xmath152 .
we will let @xmath353 denote the resonant frequency of the mode , and let @xmath354 and @xmath355 denote the corresponding wavelength and wavenumber .
we can characterize the shape of the mode by a dimensionless function @xmath62 that is given by @xmath356 where @xmath357 is the mode radius .
if we drive the cavity with an input beam that has power @xmath358 and frequency @xmath359 , then the intensity at a point @xmath11 inside the cavity is @xmath360 where @xmath361 is the detuning of the input beam from the cavity resonance , @xmath362 is the total energy decay rate for the mode , and @xmath363 , the mode volume , is given by @xmath364 in order to relate the input power to the maximum intensity inside the cavity , it is convenient to define a power @xmath365 note that @xmath249 is the same for all cavity modes ; it depends only on the cavity geometry , not the mode number .
we can then express the maximum intensity inside the cavity as @xmath366
in the raman - raman configuration , the lack of registration between the fort and raman beams causes a well - dependent distortion of the trapping potential . here
we calculate this effect . from equation ( [ eqn : h - ext - raman - raman ] ) , we see that the total potential for the raman - raman configuration is given by @xmath367 we will assume that an atom is trapped in well @xmath34 of the fort , and define a coordinate @xmath35 and a phase @xmath368 .
note that @xmath369 we will assume that the fort and raman beams drive nearby modes of the cavity , so @xmath370 . in this limit
, we can approximate @xmath371 by replacing @xmath372 with @xmath373 and replacing @xmath374 with @xmath375 : @xmath376 as in section [ sec : trapping ] , we will take the atom to be radially stationary , and take @xmath23 as a constant parameter that enters into the potential for axial motion .
we can then write the potential as @xmath377 where @xmath378 is the axial trap depth at radial coordinate @xmath23 .
the quantity @xmath379 is given by @xmath380 and @xmath263 is given by @xmath381
the fort potential is slightly weaker for the @xmath196 ground state manifold than for the @xmath197 ground state manifold , so there is a small differential stark shift . here
we calculate this effect .
the hamiltonian that describes the differential stark shift is @xmath382 where @xmath383 is the differential stark shift at an intensity maximum and @xmath384 is the maximum differential stark shift at radial position @xmath23 .
we can calculate @xmath383 as follows .
the fort depth @xmath244 is given by ( [ eqn : v - eff - cesium ] ) with @xmath385 and @xmath386 : @xmath387 thus , the differential stark shift at an intensity maximum is given by @xmath388 expanding to first order in @xmath51 , we find that @xmath389 where @xmath390 and @xmath391 , the detuning parameters at the fort wavelength @xmath253 , are given by equation ( [ eqn : detuning - parameters ] ) . | we present two schemes for driving raman transitions between the ground state hyperfine manifolds of a single atom trapped within a high - finesse optical cavity . in both schemes ,
the raman coupling is generated by standing - wave fields inside the cavity , thus circumventing the optical access limitations that free - space raman schemes must face in a cavity system .
these cavity - based raman schemes can be used to coherently manipulate both the internal and the motional degrees of freedom of the atom , and thus provide powerful tools for studying cavity quantum electrodynamics .
we give a detailed theoretical analysis of each scheme , both for a three - level atom and for a multi - level cesium atom .
in addition , we show how these raman schemes can be used to cool the axial motion of the atom to the quantum ground state , and we perform computer simulations of the cooling process . |
more than 40 years have passed since the recognition that quasars are at cosmological distances ( schmidt 1963 ; greenstein & mathews 1963 ; mathews & sandage 1963 ) and hence must be powerful energy sources .
a combination of arguments based on time variability and energetics strongly supports the view that the activity is produced by the accretion of gas onto supermassive black holes in the centers of galaxies ( e.g. salpeter 1964 ; zeldovich & novikov 1964 ; lynden - bell 1969 ) .
however , the mechanism that provides the trigger to fuel quasars remains uncertain .
recent discoveries of correlations between masses of black holes in nearby galaxies and either the mass ( magorrian et al . 1998 ) or velocity dispersion ( i.e. the @xmath4-@xmath5 relation : ferrarese & merritt 2000 ; gebhardt et al .
2000 ) of spheroids demonstrate a fundamental link between the growth of supermassive black holes and galaxy formation .
however , theoretical evidence connecting the origin of supermassive black holes and quasars to galaxy evolution remains elusive , owing to the complexity of the underlying physics and dynamics .
in particular , up to now there has been no comprehensive model to explain the origin and fueling of quasars , or their lifetimes , obscuration , demographics , self - regulation and termination , and dependence on host galaxy properties .
observations of the nearby universe suggest that rapid black hole growth may be related to massive flows of gas into the centers of galaxies .
infrared ( ir ) luminous galaxies represent the dominant population of objects above @xmath6 locally and it is believed that much of their ir emission is powered by dust reprocessing of radiation from intense nuclear starbursts ( e.g. soifer et al .
1984a , b ; sanders et al .
1986 , 1988a , b ; for a review , see e.g. soifer et al .
1987 ) . at the highest luminosities above @xmath7 , characteristic of ultraluminous infrared galaxies ( ulirgs ) , nearly all
the galaxies appear to be in advanced stages of merging ( e.g. allen et al .
1985 ; joseph & wright 1985 ; armus et al .
1987 ; kleinmann et al . 1988 ; for reviews , see sanders & mirabel 1996 and jogee 2004 ) and co observations show that they contain large quantities of gas in their nuclei ( e.g. scoville et al .
1986 ; sargent et al .
1987 , 1989 ) . some ulirgs exhibit `` warm '' ir spectral energy distributions , perhaps indicative of a buried quasar ( e.g. sanders et al .
this fact , together with the overlap between bolometric luminosities of ulirgs and quasars , led sanders et al .
( 1988a ) to propose that quasars are the descendents of an infrared luminous phase of galaxy evolution caused by mergers .
this scenario is supported by recent x - ray observations which have revealed the presence of two non - thermal point sources near the center of the ulirg ngc6240 @xcite , which are most naturally interpreted as accreting supermassive black holes that are heavily obscured at visual wavelengths ( e.g. gerssen et al .
2004 ; max et al . 2005 ) .
hydrodynamic simulations have shown that gas inflows are produced by strong gravitational torques on the gas through tidal forces during mergers involving gas - rich galaxies ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the resulting dense concentrations of gas in the inner regions of the remnant can sustain star formation at rates high enough and for sufficiently long timescales to account for many properties of ulirgs ( e.g. , * ? ? ?
? * ; * ? ? ?
* ) . however , these authors were not able to directly explore the relationship between the gas inflows and quasar activity because their simulations did not include a model for black hole growth and the impact of strong feedback from either star formation or quasars .
one of the most fundamental parameters of black hole growth is the quasar lifetime , @xmath8 , which sets the timescale for the most luminous phase of the activity .
observations generally constrain quasar lifetimes to the range @xmath9 yr ( for a review , see * ? ? ?
these estimates are primarily based on demographic or integral arguments which combine observations of the present - day population of supermassive black holes and accretion by the high - redshift quasar population ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , or incorporate quasars into models of galaxy evolution ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) or reionization of heii ( e.g. sokasian et al .
2002 , 2003 ) .
results from clustering in quasar surveys ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , the proximity effect in the ly@xmath10 forest ( bajtlik , duncan & ostriker 1988 ; haiman & cen 2002 ; yu & lu 2005 ; although such results may still be inconclusive , e.g. croft 2004 ) , and the transverse proximity effect in he ii @xcite similarly suggest lifetimes @xmath11yr .
it is not immediately obvious how quasar lifetimes @xmath11yr can be explained in the context of gas inflows triggered by galaxy mergers .
the full duration of the starburst phase is set by the timescale during which strong gas inflows are excited , which in turn is determined by the time when significant gravitational torques are exerted on the gas .
the numerical simulations of @xcite showed that this occurs for @xmath12 yr , much shorter than typical merger timescales of @xmath13 @xcite , but in good agreement with observed estimates for the gas depletion time in the central regions of ulirgs ( for a discussion , see e.g. barnes & hernquist 1992 ) .
if gas - rich mergers are indeed responsible for the origin of quasar activity , as suggested through studies of ulirgs ( e.g. sanders et al .
1988a ) and more directly from observations of quasar hosts ( e.g. stockton 1978 ; heckman et al .
1984 ; stockton & mackenty 1987 ; stockton & ridgway 1991 ; hutchings & neff 1992 ; bahcall et al . 1994 , 1995
; canalizo & stockton 2001 ) , the factor @xmath14 difference between the gas consumption time in ulirgs and quasar lifetimes must be reconciled .
semi - analytical models of supermassive black hole evolution and its correlation with galaxy structure suggest that , beyond a certain threshold , feedback energy expels nearby gas and shuts down the accretion phase @xcite .
however , these calculations provide only limits , and in neglecting the dynamics of quasar evolution they do not predict time - dependent effects such as the characteristic lifetime of the accretion phase prior to its self - termination , the fueling rates for black hole accretion , the obscuration of central sources , or the quasar light curve .
these quantities are instead usually taken to be independent input parameters of the model .
for example , the lifetime is either adopted from observational estimates or assumed to be similar to a characteristic timescale such as the dynamical time of the host galaxy disk or the @xmath15-folding time for eddington - limited black hole growth @xmath16yr for accretion with radiative efficiency @xmath17 and @xmath18 @xcite .
efforts to model quasar accretion and feedback in a more self - consistent manner ( e.g. , * ? ? ?
* ; * ? ? ?
* ) have considered the time - dependent nature of gas inflows feeding accretion and the impact of radiative heating from quasar feedback in a hydrodynamical context .
although generally restricted to one - dimensional `` toy models '' which do not include galaxy - galaxy interactions , such modeling has made important progress in predicting the characteristic duty cycles of quasars , in good agreement with e.g. , @xcite , and in demonstrating the importance of radiative feedback on the surrounding medium and quasar evolution ( for a review , see * ? ? ?
further , @xcite showed that self regulation by radiative feedback from quasars in this modeling can expel a significant quantity of gas and leave a remnant with properties similar to observed ellipticals and black hole mass corresponding to the observed @xmath19-@xmath5 relation .
similarly @xcite showed that incorporating a simple prescription for radiative heating by quasars in the context of cosmological simulations yields populations of elliptical galaxies consistent with observed optical and x - ray properties .
recently , @xcite have developed a methodology for incorporating black hole growth and feedback into hydrodynamical simulations of galaxy mergers that includes a multiphase model for star formation and pressurization of the interstellar gas by supernova feedback @xcite . using this approach ,
we have begun to explore the impact of these processes on galaxy formation and evolution .
di matteo et al .
( 2005 ) and springel et al .
( 2005b ) have shown that the gas inflows produced by gravitational torques during a merger both trigger starbursts ( as in the earlier simulations of e.g. mihos & hernquist 1994 , 1996 ) and fuel rapid black hole growth .
the growth of the black hole is determined by the gas supply and terminates abruptly when significant gas is expelled owing to the coupling between feedback energy from black hole accretion and the surrounding gas .
eventually , as the gas is heated and driven out , the remnant is no longer active because the black hole does not accrete at a high rate , leaving a dead quasar in an ordinary galaxy .
the self - regulated nature of the black hole growth explains observed correlations between black hole mass and properties of normal galaxies @xcite , as well as the color distribution of ellipticals @xcite .
these results lend support to the view that mergers have played an important role in structuring galaxies , as advocated by e.g. toomre & toomre ( 1972 ) and toomre ( 1977 ) .
moreover , @xcite showed that the dynamics of the inflowing gas and its response to the self - regulated growth of the black hole yield a timescale for the strong accretion phase @xmath12 yr , comparable to the full duration of the starburst . during much of this period
, the bolometric luminosity of the black hole would exceed the threshold to be classified as a quasar , implying a timescale @xmath20 ( for a milky way mass system ) for the intrinsic quasar phase of the black hole . in @xcite ,
we employed models for obscuration by gas and dust to show that for most of the accretion lifetime the quasars would be buried ; i.e. heavily obscured by the large gas density powering accretion . eventually , feedback from the accretion energy drives away the gas , creating a brief window in which the central object would be observable as an optical quasar , until accretion levels drop below quasar thresholds . by calculating the effects of obscuration from the simulation of a merger of gas - rich galaxies
, we showed that this determines an observable lifetime @xmath21 , in good agreement with observations . here ,
we extend and further develop implications of the model for quasar evolution proposed in @xcite by analyzing a series of hydrodynamical simulations of galaxy mergers where we vary the masses of the progenitor galaxies , so that they have virial velocities between 80 and @xmath22 . in 2 we describe our series of merger simulations and the calculation of column densities and obscuration , and compare the results obtained with different column density calculations . in 3 we examine the frequency dependence of our results and quantify the differences in lifetimes across the optical and x - ray bands . in 4 we examine the typical column densities along lines - of - sight to the simulated quasars , and compare these to observed distributions from both optical and x - ray surveys . in 5 we show the simulation results for our merger series and compare the observed and intrinsic lifetimes as a function of luminosity for all the simulations .
finally , in 6 we discuss our results and the implications of our model for a broad range of quasar studies .
the simulations were performed using the gadget-2 code , a new version of the parallel treesph code gadget @xcite .
it uses an entropy - conserving formulation @xcite of smoothed particle hydrodynamics ( sph ) , and includes radiative cooling , heating by a uv background ( as described in e.g. katz et al . 1996 ;
dav et al . 1999 ) , and a sub - resolution model of a multiphase interstellar medium ( ism ) to describe star formation and supernova feedback @xcite .
this sub - resolution model provides an effective equation of state for star - forming gas which includes pressure feedback from supernova heating , and allows us to stably evolve even massive pure gas disks ( see , e.g. springel et al .
2005b ; robertson et al .
the methodology of accretion , feedback , and galaxy generation is described in detail in @xcite . in our approach , supermassive black holes ( bhs )
are represented by `` sink '' particles that accrete gas from their local environment , with an accretion rate @xmath23 estimated from the local gas density and sound speed using a bondi - hoyle - lyttleton parameterization with an imposed upper limit equal to the eddington rate .
the bolometric luminosity of the bh particle is then @xmath24 , where @xmath25 is the radiative efficiency .
we further assume that a small fraction ( @xmath26 ) of @xmath27 couples dynamically to the surrounding gas , and that this feedback is injected into the gas as thermal energy .
this fraction is a free parameter , determined in @xcite by fitting to the @xmath19-@xmath5 relation .
we do not attempt to resolve the small - scale accretion dynamics near the black hole , but instead assume that the time - averaged accretion can be estimated from the gas properties on the scale of our spatial resolution ( @xmath28pc ) .
cccc a1 & 80 & @xmath29 & @xmath30 + a2 & 113 & @xmath31 & @xmath32 + a3 & 160 & @xmath33 & @xmath34 + a4 & 226 & @xmath35 & @xmath36 + a5 & 320 & @xmath37 & @xmath38 + in what follows , we analyze five simulations of colliding disk galaxies , which form a family of structurally similar models with different virial velocity and mass . in each simulation , we generate two stable , isolated disk galaxies , each with an extended dark matter halo with a @xcite profile , motivated by cosmological simulations ( e.g. navarro et al .
1996 ; busha et al . 2004 ) and observations of halo properties ( e.g. rines et al . 2002 , 2002 , 2003 , 2004 ) , an exponential gas disk , and a bulge .
the simulations follow the series described in detail in @xcite , with the parameters listed in table [ tbl : sims ] .
we denote the simulations a1 , a2 , a3 , a4 , and a5 , with increasing virial velocities of @xmath39 , respectively .
note that the self - similarity of these models is broken by the scale - dependent physics of cooling , star formation , and black hole accretion . in @xcite , we describe our analysis of simulation a3 , a fiducial choice with a rotation curve and mass similar to the milky way .
the galaxies have mass @xmath40 , with the baryonic disk having mass fraction @xmath41 , the bulge @xmath42 , and the rest of the mass in dark matter with a concentration parameter @xmath43 .
the disk - scale length is computed based on an assumed spin parameter @xmath44 , and the scale - length of the bulge is set to @xmath45 times this .
we begin our simulation with pure gas disks , which may better correspond to the high - redshift galaxies in which most quasars are observed . each galaxy
is initially composed of 168000 dark matter halo particles , 8000 bulge particles , 24000 gaseous disk particles , and one bh particle , with a small initial seed mass of @xmath46 . given these choices , the dark matter , gas , and star particles
are all of roughly equal mass , and central cusps in the dark matter and bulge profiles are reasonably well resolved ( see fig 2 . in springel et al .
the galaxies are then set to collide in a prograde encounter with zero orbital energy and a small pericenter separation ( @xmath47 for the a3 simulation ) .
we calculate the column density between a black hole and a hypothetical observer from the simulation outputs spaced every 10 myr before and after the merger and every 5 myr during the merger of each galaxy pair .
the calculation method is described in @xcite , but we review it briefly here .
we generate @xmath48 radial lines - of - sight ( rays ) , each with its origin at the black hole particle location and with directions uniformly spaced in solid angle @xmath49 . for each ray
, we then begin at the origin , calculate and record the local gas properties using gadget , and then move a distance along the ray @xmath50 , where @xmath51 and @xmath52 is the local sph smoothing length .
the process is repeated until a ray is sufficiently far from its origin ( @xmath53 kpc ) .
the gas properties along a given ray can then be integrated to give the line - of - sight column density and mean metallicity .
we test different values of @xmath54 and find that gas properties along a ray converge rapidly and change smoothly for @xmath55 and smaller .
we similarly test different numbers of rays and find that the distribution of line - of - sight properties converges for @xmath53 rays .
given the local gas properties , we use the gadget multiphase model of the ism described in @xcite to calculate the local mass fraction in `` hot '' ( diffuse ) and `` cold '' ( molecular and hi cloud core ) phases of dense gas and , assuming pressure equilibrium between the two phases , we obtain the local density of the hot and cold phase gas and the corresponding volume filling factors .
the values obtained correspond roughly to the fiducial values of @xcite .
given a temperature for the warm , partially ionized component @xmath56 , determined by pressure equilibrium , we further calculate the neutral fraction of this gas , typically @xmath57 .
we denote the neutral and total column densities as @xmath58 and @xmath59 , respectively .
using only the hot - phase density allows us to place an effective lower limit on the column density along a particular line of sight , as it assumes a ray passes only through the diffuse ism , with @xmath60 of the mass of the dense ism concentrated in cold - phase `` clumps . ''
given the small volume filling factor ( @xmath61 ) and cross section of such clouds , we expect that the majority of sightlines will pass only through the `` hot - phase '' component .
using @xmath24 , we model the form of the intrinsic quasar continuum sed following @xcite , based on optical through hard x - ray observations ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
this gives a b - band luminosity @xmath62 , where @xmath63 , and we take @xmath64 .
we then use a gas - to - dust ratio to determine the extinction along a given line of sight at this frequency .
observations suggest that the majority of the population of reddened quasars have reddening curves similar to that of the small magellenic cloud ( smc ; hopkins et al . 2004 ) , which has a gas - to - dust ratio lower than the milky way by approximately the same factor as its metallicity @xcite .
we therefore consider both a gas - to - dust ratio equal to that of the milky way , @xmath65 , and a gas - to - dust ratio scaled by metallicity , @xmath66 .
for both cases we use the smc - like reddening curve of @xcite .
we calculate extinction in x - ray frequencies ( 0.03 - 10 kev ) using the photoelectric absorption cross sections of @xcite and non - relativistic compton scattering cross sections , similarly scaled by metallicity . in determining the column density for photoelectric x - ray absorption ,
we ignore the inferred ionized fraction of the gas , as it is expected that the inner - shell electrons which dominate the photoelectric absorption edges will be unaffected in the temperature ranges of interest .
we do not perform a full radiative transfer calculation , and therefore do not model scattering or re - processing of radiation by dust in the infrared .
we quantify the quasar lifetime @xmath8 as in @xcite , as a function of the limiting b - band luminosity @xmath67 .
for each sightline , an observed lifetime is determined as the integrated time in the simulation during which the given sightline sees a b - band luminosity above a given threshold , @xmath68 .
we consider both the intrinsic lifetime @xmath69 , the total time the intrinsic @xmath70 ( ignoring attenuation ) , and the observed lifetime derived from a particular column density calculation .
we note that below 5 myr ( the simulation output frequency in the standard case ) our estimates of @xmath8 become uncertain owing to the effects of quasar variability and our inability to resolve the local small - scale physics of the ism .
however , the majority of sightlines see lifetimes in the range @xmath71myr up to @xmath72 , in good agreement with observations suggesting lifetimes @xmath73yr and well above this limit . from the a3 simulation .
the dotted lines show the lifetime if attenuation is ignored ( time with observed b - band luminosity @xmath74 ) .
the thick solid lines indicate the median observed lifetime , with thin lines giving the 25%-75% contours .
lifetimes are calculated using : ( a ) cold - phase ism density , ( b ) hot - phase density ( ignoring metallicity and ionization ) , ( c ) hot - phase density ( including metallicity and ionization ) , ( d ) hot - phase density with artificially low metallicity ( 0.1 times solar ) .
[ fig : comparenhcalc],width=326 ] figure [ fig : comparenhcalc ] shows the lifetime obtained in the a3 simulation using different calculations of the column density @xmath58 ( cases @xmath75 ) . the uppermost curve in all cases is the intrinsic lifetime @xmath69 .
the lower thick curve is the median observed lifetime calculated using a particular estimate of the column density , with the thin curves representing the 25%-75% inclusion contours . the cold - phase ism density ( @xmath76 ) is large enough for the quasar to be completely extincted out of observable ranges for the duration of the quasar phase , rendering the object observable only during quiescent phases with @xmath77 .
we also show the significantly longer lifetimes calculated using the hot - phase ism density , first ignoring corrections for metallicity and ionization ( @xmath78 ; i.e. assuming @xmath79 and @xmath80 ) and then including these effects ( @xmath81 ) , which increases the median lifetime by a factor @xmath82 and significantly reduces the scatter towards shorter lifetimes .
finally , we calculate the lifetime using the hot - phase density and an smc - like gas - to - dust ratio , @xmath83 ( @xmath84 ; essentially assuming a metallicity @xmath85 times solar ) , which sets a strong upper limit on the observed lifetime .
these lifetimes are still only @xmath82 times as long as the lifetimes calculated using metallicity - weighted neutral column densities , and still @xmath86 the intrinsic lifetime at @xmath72 .
we therefore expect that , after accounting for the clumping of most mass in the cold phase of the most dense regions of the ism , the qualitative relation between our observed and intrinsic calculated lifetimes should be relatively insensitive to the details of the column density calculation , with variation by a factor of @xmath87 between lifetimes calculated using different prescriptions for the multiphase ism , ionization , and metallicity effects .
we expect the intrinsic quasar lifetime above a given luminosity in a particular band to change with the frequency of that band .
the shift in lifetime with frequency is approximately given by an offset in luminosity corresponding to the difference in luminosities at different frequencies in our model quasar spectrum .
however , this change is not entirely history - independent , as the model spectrum alters shape with varying luminosity owing to bolometric corrections .
more important , our model predicts that the observed lifetime above a given luminosity will change with frequency not only as a result of the shift in intrinsic lifetime , but primarily as a result of varying levels of attenuation at different frequencies . from the a3 simulation , showing the waveband dependence of the lifetime .
the thick lines give the lifetime if attenuation is ignored , thin lines indicate the median observed lifetime ( calculated using the metallicity - weighted , hot - phase neutral column density ) .
upper panel : from black to blue to red , lifetimes are calculated at 3000 , 4500 , 6000 , 7500 , and 9000 .
lower panel : lifetimes are calculated for soft ( @xmath88 , red ) and hard ( @xmath89 , black ) x - ray bands . [
fig : lifetimevsfreq],width=336 ] figure [ fig : lifetimevsfreq ] shows the lifetimes as a function of luminosity for the a3 simulation , for various frequencies in the optical and x - ray bands .
we plot lifetimes in the manner of figure [ fig : comparenhcalc ] ; i.e. the total time with @xmath90 ( @xmath91 ) , at the representative optical wavelengths @xmath92 ( colored black to blue to red , respectively ) .
we find that the intrinsic lifetime systematically decreases at longer wavelengths ( based on the shape of the quasar sed ) , while the observed lifetime increases ( owing to weakening attenuation ) , with the two approaching one another near infrared wavelengths . at longer wavelengths
, we expect the curves to cross , as the observed lifetime in the far ir , owing to dust re - processing of quasar radiation , will be longer than the intrinsic lifetime , but we do not yet incorporate such re - processing into our column density calculations . at very low @xmath93 , lifetimes approximately converge to the total duration of the simulations , as the quasar is above such luminosities throughout almost the entire merger .
the lower panel of figure [ fig : lifetimevsfreq ] shows the lifetimes in the soft and hard x - ray bands , defined as @xmath88 and @xmath89 , respectively .
we find that the soft x - ray band is heavily attenuated and has much shorter observed than intrinsic lifetimes , primarily owing to photoelectric absorption .
however , the hard x - ray band is relatively unaffected , as the hot - phase column densities tend to be well below the compton - thick regime @xmath94 .
thus , most quasars obscured in the optical for much of their high accretion rate lifetimes should be observable in hard x - rays .
we defer a full treatment of the difference between the attenuation of soft and hard x - rays to a future paper , but note that this difference may account for the slope of the cosmic x - ray background , with a population of obscured quasars at spectral energies @xmath95 a few kev as a natural stage in the evolution of quasar activity .
given a calculation of the column density along multiple lines - of - sight to the simulated quasars , we can examine the typical column density distributions at all times and at times in which the quasar is observed above some luminosity threshold .
following the analysis in @xcite , we expect the characteristic column densities to rise rapidly during the merger , as gas inflow traces and fuels a rising accretion rate .
eventually , column densities rapidly fall in a blowout phase once the black hole has reached a critical mass , creating a window with rapidly changing column densities in which the quasar is observable until accretion rates drop below those necessary to fuel quasar activity .
as is clear in figure 2 of @xcite , the dispersion in @xmath59 at any particular time in the simulation is generally small , a factor of @xmath82 in either direction , but typical @xmath59 values can change by an order of magnitude over timescales @xmath96myr .
therefore , it is of interest to consider the probability of an observer , viewing the object at a random time , seeing a given column density . from the a3 simulation .
the probability of seeing a given @xmath58 is weighted by the time such an @xmath58 is observed along all sightlines .
plots are scaled linearly ( left ) and logarithmically ( right ) .
the top panel shows all @xmath58 values from all times , the lower panels give the distribution for times where the observed b - band luminosity is above the @xmath67 shown .
the smooth curve is the distribution of @xmath58 values from the observed sdss quasar sample of @xcite , who used an absolute @xmath97-band magnitude limit approximately equivalent to @xmath98 , scaled to the appropriate modal @xmath58 .
[ fig : nhdistrib],width=355 ] figure [ fig : nhdistrib ] plots the distribution of column densities @xmath58 calculated using the hot - phase metallicity - weighted ism density in the a3 simulation . the probability of seeing a given @xmath58 is calculated as proportional to the total time along all sightlines that such an @xmath58 is observed .
the plots are scaled both linearly ( left panels ) and logarithmically ( right panels ) in column density .
typical column densities are distributed about @xmath99 , approximately symmetrically in @xmath100 .
we further plot the distribution of @xmath58 values requiring that the observed b - band luminosity be above some threshold @xmath67 .
the curve shown in the linear plots is the best - fit to the @xmath101 distribution of bright sdss quasars with @xmath102 , from @xcite .
the curve has been rescaled in terms of the column density ( inverting our gas - to - dust prescription ) and plotted about a peak ( mode ) @xmath58 ( undetermined in hopkins et al .
2004 ) of @xmath103 .
the @xmath97-band absolute magnitude limit imposed in the observed sample , @xmath104 , corresponds approximately to our plotted b - band limit @xmath98 .
the agreement between the observed column density distribution and the result of our simulations once the same selection effect is applied is strong evidence in favor of our model for quasar evolution .
the distribution as a function of limiting luminosity is a natural consequence of the dynamics of the quasar activity . throughout much of the duration of bright quasar activity
, column densities rise to high levels as a result of the same process that feeds accretion , naturally producing a reddened population of quasars ( a red `` tail '' in the quasar color distribution ) , extending to very bright quasars strongly reddened by large @xmath58 .
the existence of this extended reddened quasar population in radio and optically selected quasars is well known ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , and we are able to reproduce its distribution not as a distribution of source properties but as a result of the time evolution of quasar phenomena .
our estimate of the distribution of @xmath58 for @xmath98 does not account for an additional selection effect , namely that strongly reddened quasars may not be extincted from an observed sample ( if the intrinsic luminosity is large enough ) , but their colors may be significantly reddened to the point where color selection criteria of quasar surveys will not include them .
however , this effect would only serve to bring our distribution into better agreement with observations , as it would slightly lower the high-@xmath58 tail .
moreover , our estimates of the distribution of @xmath58 allow us to make a directly observable prediction , namely that the column density distribution in optically selected samples with a minimum absolute magnitude or luminosity should broaden in both directions ( to larger and smaller @xmath58 ) as the limiting selection luminosity is decreased ( limiting absolute magnitude increased ) .
this is because , at lower luminosities , observers will see both intrinsically bright periods extincted by larger column densities ( broadening the distribution to larger @xmath58 values ) and intrinsically faint periods with small column densities ( broadening the distribution to smaller @xmath58 values ) .
the distribution of @xmath59 values in hard x - ray quasar samples , which are much less affected by this extinction ( see 3 ) , should also reflect this trend , and indeed the distribution of @xmath59 values is much broader and flatter , with quasars seen from @xmath105 @xcite .
extension of this distribution to the other black hole masses and peak quasar luminosities of our simulations reveals similar qualitative behavior .
the characteristic column densities during the obscured , intrinsically bright phases of accretion generally increase with galaxy and black hole mass , but typically by factors @xmath106 across the range of our simulations @xmath107 . additionally ,
although mean column densities increase somewhat with increasing mass , the functional form of the @xmath58distribution remains similar at each limiting luminosity , implying that a combined population of quasars similar to those we have simulated will match the observed @xmath58 distribution with appropriate selection effects accounted for as in figure [ fig : nhdistrib ] .
finally , we note that although we do not see extremely large column densities @xmath108 in the distributions from our simulations , our model does not rule out such values .
it is possible that very bright quasars in unusually massive galaxies or quasars in higher - redshift , compact galaxies which we have not simulated may , during peak accretion periods , reach such values in their typical column densities .
moreover , as our model assumes @xmath109 of the mass of the densest gas is clumped into cold - phase molecular clouds , a small fraction of sightlines will pass through such clouds and encounter column densities similar to those shown for the cold phase in figure [ fig : comparenhcalc ] , as large as @xmath110 .
this also allows a large concentration of mass in sub - resolution obscuring structures , such as an obscuring toroid on scales @xmath111pc , although many of the phenomena such structures are invoked to explain can be accounted for through our model of time - dependent obscuration .
as a function of simulation time , with thin contours at 25% and 75% inclusion levels , for the a4 simulation ( @xmath112 ) .
solid contours represent the density of the `` hot - phase '' ism , dashed contours the total simulation density .
middle panel : bolometric luminosity of the black hole in the simulation , @xmath24 ( thick ) , and ratio of bolometric to eddington luminosity , @xmath113 ( thin ) .
values are shown for each simulation timestep .
lower panel : observed b - band luminosity with @xmath58 calculated using the median `` hot - phase '' ism density . [ fig : nh.vs.time.a4],width=336 ] , but for the a5 ( @xmath114 ) simulation .
[ fig : nh.vs.time.a5],width=336 ] we generalize the results from @xcite and those above ( which use simulation a3 ) to a range of masses , using the simulations a1 , a2 , a4 , and a5 .
recall , these simulations are identical except for increasing galaxy mass given by the sequence in virial velocity , @xmath115 for a1-a5 , respectively .
the evolution of these simulations would be simply related if not for the scale - dependent physics of cooling , star formation , and black hole accretion and feedback .
figures [ fig : nh.vs.time.a4 ] and [ fig : nh.vs.time.a5 ] plot the bolometric luminosity of and column densities to the supermassive black hole as a function of simulation time , for the period shortly before and after the merger in simulations a4 and a5 , respectively .
this is presented in the style of figure 2 from hopkins et al .
( 2005a ) , which showed the same for a3 .
the median column densities and 25%-75% contours are indicated for the total average simulation density , and the calculated hot - phase density , as described in 2.2 , along with the observed b - band luminosity using the median hot - phase @xmath58 .
the final masses of the black holes are also shown , where @xmath116 for simulations a1-a5 , respectively .
the evolution seen in simulation a4 is quite similar to that of a3 , with a sharp rise at the time of the merger to very large peak eddington - limited luminosities , for a period @xmath117 , much of which is unobservable in the optical owing to a corresponding rise in @xmath58 . at the final stages of strong accretion ,
when gas is rapidly being consumed and expelled , the observable luminosity finally rises to peak levels over a much shorter time interval @xmath118 . in the a5 simulation , the period of rapid accretion and large , eddington - limited intrinsic luminosities is significantly longer , @xmath119 , and the peak luminosity is higher , as expected .
however , as before , most of this period is obscured by large column densities , and , remarkably , once again only the final stages of intrinsically high luminosity are visible , as gas is expelled and the accretion rate and column density both fall .
the same behavior is seen in the a1 and a2 cases , though with much lower peak bolometric luminosities @xmath120 and @xmath121 , respectively .
the striking similarity of this process across all our simulations suggests that it is at least qualitatively insensitive to the details of the accretion history and galaxy or black hole masses . from the simulations ( time with observed b - band luminosity @xmath122 ) .
results from simulations a1-a5 are shown from top to bottom .
solid lines indicate the lifetime if attenuation is ignored .
the dashed lines give the median observed lifetime calculated with the metallicity - weighted , hot - phase neutral column density .
right panels : same as the left panels , but with @xmath8 defined as a bolometric luminosity @xmath123 , where the observable lifetime is the time above a given ratio @xmath124 with b - band optical depth less than unity .
[ fig : lifetime.vs.l],width=336 ] we quantify the resulting quasar lifetimes in figure [ fig : lifetime.vs.l ] , where we plot the quasar lifetime @xmath8as a function of the limiting b - band luminosity @xmath67 .
the panels show the lifetimes for simulations a1-a5 , from top to bottom . in each panel
, the solid curve shows the intrinsic lifetime @xmath69 ; i.e. the total time the intrinsic @xmath70 .
the dashed curves show the integrated time that the observed b - band luminosity meets this criterion , using the median metallicity - weighted hot - phase neutral column density of the simulation ( case @xmath81 of figure [ fig : comparenhcalc ] ) .
as demonstrated in figure [ fig : comparenhcalc ] and in @xcite , varying this definition of the column density over realistic ranges can change the intrinsic lifetime by a factor of @xmath87 at a given @xmath67 .
we also plot the lifetime as a function of @xmath124 , the ratio of the bolometric luminosity to the eddington luminosity , and find a similar trend . here
, we calculate the `` observed '' time above a given ratio @xmath124 as the time above such a ratio with a b - band optical depth less than unity . in all cases ,
the observed @xmath8 is significantly smaller than the intrinsic lifetime for all @xmath125 , and the ratio @xmath126 decreases with increasing @xmath67 .
however , both the intrinsic and observed lifetimes increase systematically at a given @xmath127 with increasing galaxy and black hole mass , and the peak luminosity similarly increases , as expected . in all cases corresponding to quasar - like luminosities ,
our estimates of observed lifetimes of @xmath118 agree well with observations , with much longer intrinsic lifetimes @xmath128 . a quasar accreting and radiating at the eddington limit , with luminosity increasing exponentially up to some peak luminosity @xmath129 after which accretion shuts off , will spend an integrated time above a given luminosity @xmath130 given by @xmath131 where @xmath132 is the salpeter time , @xmath133yr at the eddington rate with our efficiency @xmath25 , and @xmath129 is given by the eddington luminosity of the final black hole mass .
this lifetime agrees well with the plotted lifetimes for simulations a3 , a4 , and a5 at the high - luminosity end .
however , simulations a1 and a2 do not radiate at their eddington limits for a long period of time , and the slope of this relation is too shallow in all cases , underpredicting the time that the simulation spends at luminosities @xmath134 .
this is because the black holes in the simulations spend significant time at lower eddington ratios both going into and coming out of the quasar stage and in an extended quiescent phase .
moreover , the observed lifetimes do not at all correspond to the lifetimes predicted by this simple model .
therefore , to simply describe the intrinsic and observed lifetimes as a function of limiting luminosity and final black hole mass ( or peak luminosity @xmath129 ) , we fit the calculated lifetimes @xmath8 to truncated power laws , from @xmath135 to approximately @xmath136 , or where @xmath137myr , whichever occurs first .
thus we fit to the form @xmath138 where @xmath139 and @xmath140 . to truncated power laws of the form @xmath141 .
upper panel shows the intrinsic ( thick solid line ) and observed ( thin solid line ) lifetimes in the a3 simulation , with dashed lines indicating the corresponding best - fit truncated power laws .
the dotted line shows the lifetime expected for eddington - limited accretion up to the final black hole mass .
the lower panels plot the power law slopes @xmath10 and normalizations at @xmath135 of the power law fits to the intrinsic ( thick ) and observed ( thin ) lifetimes from the series of simulations a1-a5 , as a function of the final black hole mass in the simulation .
the dotted line in the left panel gives a simple linear fit to @xmath10 as a function of the logarithm of the final black hole mass .
[ fig : pwrlawfits],width=336 ] figure [ fig : pwrlawfits ] shows the resulting power - law slopes @xmath10 and normalizations @xmath142 of the lifetimes @xmath8 as a function of the final black hole mass ( or equivalently , the peak quasar luminosity given as the eddington luminosity of the final black hole mass ) .
the calculated , fitted , and eddington - limit approximated lifetimes are also shown for comparison for the a3 simulation .
the lifetimes are well - fitted by power laws in this range , and the discrepancy between the power - law and eddington - limit approximated lifetimes at lower luminosities is clear .
the slope of both the intrinsic and observed lifetimes becomes shallower at higher black hole masses , as the lifetime increases for all luminosities and extends to higher peak luminosities , but the slope of the observed lifetime evolves much more rapidly than the slope of the intrinsic lifetime . over the range of our simulations ,
the intrinsic lifetime slope can be approximated as @xmath143 , or @xmath144 .
although the fluctuation in the normalization @xmath142 is large for observed lifetimes , there does not appear to be a systematic trend with final black hole mass , which is expected as by luminosities of this level or lower , the lifetime is dominated by the quiescent - phase lifetime throughout the duration of the simulation .
this is especially clear in the normalization of the intrinsic lifetimes , which also show no systematic trend and are all within a factor of @xmath145 two of 2 gyr , the approximate total duration of the simulations . in order to obtain a lower limit to quasar lifetimes at low luminosities , we repeat this analysis , but ignore all times prior to the intrinsically bright quasar phase ( essentially the final merger )
this excludes the luminosity of the black hole as it grows from its small seed mass , and ignores the periods of accretion with high gas fractions early in the merger , giving a lower limit corresponding to an already - large black hole suddenly `` turning on '' in a quasar phase .
as expected , the eddington - limited phase of the lifetime is more dominant in this case .
however , the pure eddington - limited model still gives lifetimes too short by an order of magnitude at low luminosities .
fitting to power laws we find very similar relations to those shown in figure [ fig : pwrlawfits ] , with slightly shallower slopes , @xmath146 for the intrinsic lifetime , and normalization @xmath147gyr .
we further fit one additional functional form to the intrinsic lifetimes , in which the lifetime follows the eddington - limited shape for large luminosities , down to some luminosity @xmath148 , and below this luminosity follows a power - law .
this allows a continuous description of the lifetime , very accurate at both low and near - peak luminosities . for these fits
, we find similar power - law slopes @xmath149 , with @xmath150 nearly constant .
we also fit the lifetimes as a function of frequency for the representative visual wavelengths shown in figure [ fig : lifetimevsfreq ] .
although , as expected from the figure , the observed lifetime slopes become slightly shallower and the intrinsic lifetime slopes become slightly steeper as the quasar is observed at redder wavelengths , the change with wavelength across this range is very small compared to the change with black hole mass .
moreover , the change in normalization @xmath142 across visual wavelengths is consistent with zero .
we therefore find that quasar lifetimes can be well approximated by simple power laws , where the power law slope depends on the final black hole mass or peak luminosity of a system , and find significantly longer lifetimes at low luminosities than predicted by simple models of accretion at a constant eddington ratio .
furthermore , attempts to de - convolve intrinsic quasar properties and observations as well as semi - analytical attempts to model observable quasar luminosity functions and statistics require an accurate model of quasar lifetimes as a function of luminosity and host system properties . the fits
we describe have the advantage of a simple , analytical form which can be applied to a continuum of luminosities and intrinsic quasar properties , while addressing the shortcomings of simpler lifetime models below peak system luminosities .
in this paper , we have studied implications of the conclusion of @xcitethat incorporating the effects of obscuration in a galaxy merger simulation gives observed quasar lifetimes of @xmath118 .
we extended our analysis to a series of five simulations with virial velocities @xmath151 , producing a range of final black hole masses from @xmath152 , and find qualitatively similar results in all cases and for different determinations of the obscuring column density .
the good agreement with observations and the significant difference between observed and much longer intrinsic lifetimes is found in all our simulations .
the qualitative evolution of quasars in this model is robust , with the result reproduced across a range of galaxy masses , and using different methods for calculating column densities that can give factors greater than an order of magnitude different attenuations along any given line of sight at a particular time .
we therefore expect that quasars will have extended intrinsic lifetimes , much of which are obscured in visual wavelengths .
the processes which fuel periods of quasar activity by channeling significant quantities of gas into the central regions of a galaxy will produce this extended intrinsic lifetime of @xmath153yr .
however , for a significant fraction of the intrinsic lifetime for quasar activity , this same process will produce large column densities which heavily obscure the quasar and attenuate it well below observable limits in the b - band and other visual wavelengths .
eventually , feedback from accretion energy will remove surrounding gas , creating a window in which the quasar is optically observable before accretion rates drop below those needed to maintain quasar luminosities . of course
, not all aspects of agn activity are related to mergers .
for example , some low redshift quasars ( e.g. bahcall et al . 1996 ) and many seyferts appear to reside in relatively ordinary galaxies .
gas expelled by quasar feedback may cool and relax to the central regions of the galaxy , and passive stellar evolution can produce a large fuel supply in winds . in our picture , it is most natural to interpret occurrences of agn activity that are not caused by mergers as arising from normal galaxies being re - activated as gas is sporadically accreted by a black hole created earlier in a bright quasar phase .
the principle requirement of this modeling is that such activity should not contribute a large fraction of the black hole mass , to avoid spoiling tight correlations between the black hole and host galaxy properties or overproducing the present - day density of very massive black holes .
radiative heating from the final stages of quasar activity may suppress much of this activity ( much of the remaining gas is heated to very high temperatures ; e.g. cox et al .
2005 ) , and the hydrodynamical modeling of @xcite demonstrates that residual accretion from stellar feedback and `` cooling flows '' is dramatically suppressed by even relatively small agn feedback once the black hole has reached a critical mass determined by the @xmath19-@xmath5 relation .
further , demographic arguments such as that of @xcite , as well as more detailed analysis by e.g. , @xcite implies that most of the present black hole mass density is accumulated in bright quasar phases , for which our modeling should be most applicable .
this remains true in our analysis , and we demonstrate in hopkins et al .
( 2005b , c ) that our modeling can account for the entire quasar luminosity function in different frequencies , although we defer a detailed prediction of the resulting supermassive black hole density and mass function to a future paper .
it is important to note , however , that given our prediction of an obscured accretion phase with a duration up to @xmath154 times that of the optically observable phase , a naive application of the @xcite or other demographic arguments would imply a similar increase in the black hole mass density accumulated in obscured quasar phases .
but this neglects both the fact that black holes are growing during this phase and the luminosity dependence of the quasar lifetime .
a more detailed calculation shows that a large fraction ( @xmath155 in the brightest quasars ) of the quasar mass is accumulated in the optically observable stages of black hole growth when the final @xmath15-folding of exponential black hole growth occurs as the black hole begins to drive strong feedback and soon shuts down accretion .
similarly , most of the total black hole radiated energy is emitted in this stage , with a considerable fraction ( @xmath156 ) observable .
thus , we expect that corrections to the estimates of black hole growth from demographic arguments based on observations of bright , relatively unobscured quasars will in fact only be of order unity from our modeling , and thus should not conflict with theoretical limits .
it is still an interesting and important question , however , to determine from our modeling the implied relic agn density and constraints from the combination of observed luminosity functions and black hole mass functions .
we do not expect other considerations , such as the orbital parameters of the merger or the collimation of black hole feedback , to substantially change our results .
different orbital separations , orientations , and energies will change the time of interaction , but ultimately our model for quasar evolution depends only on the merger eventually depositing sufficient gas into the central regions of the remnant to fuel rapid black hole growth .
after such a process begins , the evolution of the black hole should be determined by the self - regulated mechanisms which terminate black hole growth , as described in @xcite .
this will eventually expel gas , rendering the quasar observable and shutting down the accretion phase at the final black hole mass given by the @xmath19-@xmath5 relation .
the phase of rapid black hole growth and the expulsion of gas in the final stages of strong accretion occur during the end - stages of the merger and in the very center of the merging cores , when and where the structure of the original orbits should be least important .
this view is supported by additional simulations not presented here but shown in e.g. fig .
16 of @xcite , demonstrating that the final black hole mass is relatively insensitive to orbital parameters .
the exact mechanism of black hole feedback should similarly not dramatically change this picture , so long as a period of rapid black hole growth occurs , requiring large densities ( which naturally generate obscuring column densities ) , and ultimately terminates via some self - regulated mechanism which expels gas from central regions , rendering the quasar optically observable in the final stages of its life .
in particular , collimation of the black hole feedback may significantly increase the energy or momentum input on - axis , `` blowing out '' material along this axis earlier in the quasar lifetime .
however , since the black hole is growing exponentially during peak accretion , this should occur at most on the order of a couple of salpeter times ( during which the black hole luminosity grows by an order of magnitude ) before it would in the isotropic case , changing the intrinsic lifetime by only a fraction of its total .
the unobscured stage of the quasar lifetime may then begin slightly earlier along on - axis sightlines , but should be of comparable duration as a similar mechanism removes the gas from central regions .
in addition , in order for significant black hole growth to cease at the same critical mass as in the isotropic case ( determined by the @xmath4-@xmath5 relation ) , the high - accretion rate phase must end over the same timescale @xmath157 as the `` lead '' with which collimated , on - axis blowout precedes isotropic blowout .
the necessity that self - regulation eventually terminates accretion implies that surrounding densities , even off - axis , must decrease , resulting in at most a broadening of the column density distribution as typical column densities off - axis could be somewhat larger than column densities on - axis . for much of this regime
, we find that the source should be observable in hard x - ray frequencies , as the attenuation is weaker and column densities generally lie below the compton thick limit .
this distinction naturally produces a substantial population of obscured quasars , as a standard phase in the evolution of quasars over their lifetimes .
although we defer a detailed calculation of the quasar luminosity function as a function of redshift and in different observed frequencies ( hopkins et al .
2005b ) , we note that such a population can naturally account for both the cosmic x - ray background spectrum and discrepancies between optical and x - ray luminosity functions at various redshifts ( hopkins et al .
moreover , optical and x - ray observational samples will be affected rather differently by selection effects and magnitude limits , especially for quasars near their peak luminosities .
therefore , the differences in the evolution of optical and x - ray selected samples may be accounted for through the dependence of obscuration on intrinsic luminosity and host galaxy properties .
attenuation is also weaker in the infrared , and the large fraction of optical and uv energy absorbed during the obscured phase may be reprocessed by dust , appearing as thermal radiation in the ir .
this may render the observable lifetime larger than the intrinsic lifetime at long wavelengths , an effect important for calculating the ir background and estimating source populations .
similarly , the total obscuration and obscuration as a function of intrinsic luminosity are important for the evolution of luminous and ultraluminous infrared galaxies , and for estimating the relative energy contributions in the ir of starbursts and active galactic nuclei ( agn ) .
the typical column densities in our simulations correspond well to observed column densities of optically selected quasars , once the appropriate observed magnitude limit has been imposed .
thus , our model allows us to reproduce the distribution of optically obscured sources above this magnitude limit naturally , as an evolutionary effect of the mechanisms which fuel and regulate quasar growth , without invoking particular distributions of source properties or geometric patterns of obscuration .
moreover , the column densities we calculate in our simulations allow us to predict that the distribution of @xmath59 values should become broader as the minimum observable luminosity is decreased , as both faint , quiescent phases with low ( @xmath158 ) column densities and bright , obscured phases with high ( @xmath159 ) column densities become observable .
this effect is seen in the broad distribution of @xmath59 values in x - ray samples , which are much less affected by attenuation , but a more detailed analysis including modeling the distribution of quasar properties is needed to reproduce both the optical and x - ray @xmath59 distributions more accurately .
we find that the peak luminosities and lifetime above any given luminosity increase systematically with galaxy mass and final black hole mass , although the distinction between observed and intrinsic lifetimes remains significant in all cases .
moreover , the ratio of observed to intrinsic lifetimes decreases in all cases with increasing minimum luminosity .
we also find that intrinsic lifetimes at luminosities @xmath160 times the peak luminosity are poorly fit by assuming quasars always accrete near the eddington limit , but rather that lifetimes are well - fitted by power laws with a steeper slope .
this is a result of sub - eddington accretion rates both before and after the peak accretion phase , even at luminosities significantly above the quiescent steady - state small accretion rates seen at late times in the simulations .
this suggests that many quasars seen at low luminosities may be quasars with a large peak luminosity in a significantly sub - eddington phase , although the fraction of quasars observed at high eddington ratios becomes large with increasing luminosity .
additionally , these lifetimes imply that , even at high luminosities where growth may be eddington - limited , quasars spend a significant fraction of their lives with intrinsic luminosities well below their peak luminosities . therefore , any observed luminosity function is the convolution of the distribution of quasars with a given peak luminosity ( determined by the final black hole mass and thus the merging galaxy properties ) and a non - trivial light curve ( hopkins et al .
it is clear from these calculations that any attempt to theoretically model even the intrinsic luminosity functions of quasars must take into account the functional dependence of the light curve on luminosity and time . in order to apply these models to observed luminosity functions , the dependence of observed luminosity on intrinsic luminosity and the quasar evolution , as well as the observed frequency , must be considered ( hopkins et al . 2005b , c ) .
we find that the ratio of observable to intrinsic lifetimes is a strongly decreasing function of the limiting luminosity of observations .
modeling this effect is necessary to estimate intrinsic quasar lifetimes from observations , as well as for using theoretically motivated accretion models to predict the quasar luminosity function and space density of present - day supermassive black holes .
further , the effects we describe can account for the presence of an obscured population of quasars which are missed by optical , uv , or soft x - ray surveys but may contribute significantly to the cosmic x - ray background .
observations of the cosmic x - ray background and comparison of the optical and hard x - ray quasar luminosity functions that indicate the existence of a large obscured population of quasars ( * ? ? ?
* and references therein ) , is explained in our picture , because it predicts different observed lifetimes and populations at different frequencies .
the scenario we describe also implies that the reprocessing of quasar radiation by dust in surrounding regions can account for observations of luminous and ultraluminous infrared galaxies with merger activity and obscured agn .
together with the modeling presented by di matteo et al .
( 2005 ) , springel et al .
( 2005a , b ) , springel & hernquist ( 2005 ) , hopkins et al .
( 2005a , b , c ) , and robertson et al .
( 2005 ) , the results described here motivate the following picture for galaxy formation and evolution . through the hierarchical growth of structure in a cold dark matter universe ( white & rees 1978 )
, mergers between galaxies occur on a regular basis .
those involving gas - rich progenitors , which would be increasingly more common towards higher redshifts , produce inflows of gas through gravitational torques @xcite , causing starbursts @xcite like those associated with luminous infrared galaxies ( e.g. sanders & mirabel 1996 ) .
the high gas densities triggering these starbursts fuel rapid black hole growth . for most of the period over which black hole growth
occurs , optical quasar activity would be buried , but x - rays from the black holes explain the presence of non - thermal point sources in e.g. ngc 6240 @xcite , and reprocessing of most of the black hole energy by surrounding gas and dust can , in principle , account for the spectral energy distributions of `` warm '' ulirgs ( sanders et al .
1988c ) and recent observations of a correlation between quasar obscuration and far - infrared host luminosity ( e.g. , * ? ?
* ; * ? ? ?
as the black hole mass and radiative output increase , a critical point is reached where feedback energy starts to expel the gas fueling accretion . for a relatively brief period of time ,
the galaxy would be seen as an optical quasar with a b - band luminosity and lifetime characteristic of observed quasars .
this phase of evolution is brief ( @xmath161 yr ) , owing to the explosive nature of the final stages of black hole growth as the gas responds dramatically to the feedback energy from the exponentially evolving black hole .
this agn feedback terminates further black hole growth , leaving a remnant that resembles an ordinary galaxy containing a dead quasar and satisfying the @xmath19-@xmath5 relation .
modeling both the dependence of quasar lifetime on luminosity , and the complex , time - dependent evolution of quasar obscuration is thus crucial to any observational or theoretical understanding of quasars and quasar host galaxy evolution .
this work was supported in part by nsf grants aci 96 - 19019 , ast 00 - 71019 , ast 02 - 06299 , and ast 03 - 07690 , and nasa atp grants nag5 - 12140 , nag5 - 13292 , and nag5 - 13381 .
the simulations were performed at the center for parallel astrophysical computing at the harvard - smithsonian center for astrophysics .
allen , d.a . ,
roche , p.f .
& norris , r.p . 1985 , mnras , 213 , 67p armus , l. , heckman , t.m . & miley , g. 1987 , , 94 , 831 bahcall , j.n . , kirhakos , s. & schneider , d.p .
1994 , , 435 , l11 bahcall , j.n . , kirhakos , s. & schneider , d.p .
1995 , , 450 , 486 bahcall , j.n
. , kirhakos , s. & schneider , d.p .
1996 , , 457 , 557 bajtlik , s. , duncan , r. c. , & ostriker , j. p. 1988
, , 327 , 570 barnes , j. e. 1988 , , 331 , 699 barnes , j. e. 1992 , , 393 , 484 barnes , j. e. & hernquist , l. 1991 , , 370 , l65 barnes , j. e. & hernquist , l. 1992 , , 30 , 705 barnes , j. e. & hernquist , l. 1996 , , 471 , 115 bouchet , p. , lequeux , j. , maurice , e. , prevot , l. , & prevot - burnichon , m. l. 1985 , , 149 , 330 brandt , w.n .
, hasinger g. , 2005 , ara&a , in press brotherton , m. s. , tran , h. d. , becker , r. h. , gregg , m. d. , laurent - muehleissen , s. a. , @xmath162 white , r. l. 2001 , , 546 , 775 busha , m.t . , evrard , a.e . ,
adams , f.c . & wechsler , r.h .
2004 , mnras , submitted [ astro - ph/0412161 ] canalizo , g. & stockton , a. 2001 , , 555,719 ciotti , l. & ostriker , j. p. 1997
, , 487 , l105 ciotti , l. & ostriker , j. p. 2001
, , 551 , 131 cox , t. j. , di matteo , t. , hernquist , l. , hopkins , p. f. , robertson , b. , & springel , v. 2005 , , submitted [ astro - ph/0504156 ] croft , r. a. c. 2004 , , 610 , 642 dav , r. , hernquist , l. , katz , n. & weinberg , d.h . 1999 ,
apj , 511 , 521 di matteo , t. , croft , r. a. c , springel , v. , & hernquist , l. 2003 , , 593 , 56 di matteo , t. , croft , r. a. c , springel , v. , & hernquist , l. 2004 , , 610 , 80 di matteo , t. , springel , v. , & hernquist , l. 2005 , nature , 433 , 604 elvis , m. , et al .
1994 , , 95 , 1 fabian , a. c. 1999 , mnras , 308 , l39 ferrarese , l. & merritt , d. 2000 , , 539 , l9 francis , p. j. , drake , c. l. , whiting , m. t. , drinkwater , m. j. , @xmath162 webster , r. l. 2001 , publications of the astronomical society of australia , 18 , 221 gebhardt , k. , bender , r. , bower , g. et al .
2000 , , 539 , l13 george , i. m. , turner , t. j. , netzer , h. , nandra , k. , mushotzsky , r. f. , & yaqoob , t. 1998 , , 114 , 73 gerssen , j. et al .
2004 , , 127 , 75 granato g.l . , de zotti g. , silva l. , bressan a. , danese l. , 2004 , , 600 , 580 grazian , a. , negrello , m. , moscardini , l. , cristiani , s. , haehnelt , m. g. , matarrese , s. , omizzolo , a. , & vanzella , e. 2004 , , 127 , 592 greenstein , j.l . & mathews , t.a .
1963 , nature , 197 , 1041 gregg , m. d. , lacy , m. , white , r. l. , glikman , e. , helfand , d. , becker , r. h. , @xmath162 brotherton , m. s. 2002 , , 564 , 133 haehnelt , m. g. , natarajan , p. , & rees , m. j. 1998 , , 300 , 817 haiman , z. & cen , r. 2002 , , 578 , 702 haiman , z. , ciotti , l. , & ostriker , j. p. 2004
, , 606 , 763 heckman , t. m. , bothun , g. d. , balick , b. , & smith , e. p. 1984
, , 89 , 958 hernquist , l. 1989 , nature , 340 , 687 hernquist , l. 1990 , , 356 , 359 hernquist , l. 1992 , , 400 , 460 hernquist , l. 1993 , , 409 , 548 hernquist , l. & mihos , j.c .
1995 , , 448 , 41 hopkins , p. f. , hernquist , l. , martini , p. , cox , t. j. , robertson , b. , di matteo , t. , & springel , v. 2005a , , accepted [ astro - ph/0502241 ] hopkins , p. f. , hernquist , l. , cox , t. j. , robertson , b. , di matteo , t. , & springel , v. 2005b , , submitted [ astro - ph/0504252 ] hopkins , p. f. , hernquist , l. , cox , t. j. , robertson , b. , di matteo , t. , & springel , v. 2005c , , submitted [ astro - ph/0504253 ] hopkins , p. f. , strauss , m. a. , hall , p. b. , richards , g. t. , cooper , a. s. , schneider , d. p. , vanden berk , d. e. , jester , s. , brinkmann , j. , & szokoly , g. p. 2004
, , 128 , 1112 hutchings , j. b. , & neff , s. g. 1992 , , 104 , 1 jakobsen , p. , jansen , r. a. , wagner , s. , & reimers , d. 2003 , , 397 , 891 jogee , s. 2004 , in agn physics on all scales , in press [ astro - ph/0408383 ] joseph , r.d .
& wright , g.s .
1985 , mnras , 214 , 87 katz , n. , weinberg , d.h . & hernquist , l. 1996 , apjs , 105 , 19 kauffmann , g. & haehnelt , m. 2000 , , 311 , 576 kawata , d. , & gibson , b. k. 2005 , , 358 , l16 kleinmann , s.g .
1988 , , 328 , 161 komossa , s. , burwitz , v. , hasinger , g. , predehl , p. , kaastra , j. s. , & ikebe , y. 2003 , , 582 , l15 lynden - bell , d. 1969 , nature , 223 , 690 magorrian , j. et al .
1998 , , 115 , 2285 marconi , a. , risaliti , g. , gilli , r. , hunt , l. k. , maiolino , r. , & salvati , m. 2004 , , 351 , 169 martini , p. 2004 , in carnegie obs .
1 , coevolution of black holes and galaxies , ed . l.c .
ho ( cambridge : cambridge univ . press ) , 170 mathews , t.a . & sandage , a.r .
1963 , , 138 , 30 max , c.e .
2005 , , in press [ astro - ph/0411590 ] mckee , c. f. & ostriker , j. p. 1977
, , 218 , 148 mihos , j. c. & hernquist , l. 1994 , , 437 , 611 mihos , j. c. & hernquist , l. 1996 , , 464 , 641 morrison , r. & mccammon , d. 1983 , , 270 , 119 j. f. , frenk c. s. , white s. d. m. , 1996 , , 462 , 563 ostriker , j. p. , & ciotti , l. 2005 , royal society of london philosophical transactions series a , 363 , 667 page , m. j. , stevens , j. a. , ivison , r. j. , & carrera , f. j. 2004 , , 611 , l85 pei , y. c. 1992 , , 395 , 130 perola , g. c. , matt , g. , cappi , m. , fiore , f. , guainazzi , m. , maraschi , l. , petrucci , p. o. , & piro , l. 2002 , , 389 , 802 porciani , c. , magliocchetti , m. , & norberg , p. 2004
, , 355 , 1010 richards , g. t. , fan , x. , schneider , d. p. , vanden berk , d. e. , strauss , m. a. , york , d. g. , anderson , j. e. , anderson , s. f. , et al .
2001 , , 121 , 2308 k. , geller m. j. , diaferio a. , kurtz m. j. , jarrett t. h. , 2004 , , 128 , 1078 k. , geller m. j. , diaferio a. , mahdavi a. , mohr j. j. , wegner g. , 2002 , , 124 , 1266 k. , geller m. j. , diaferio a. , mohr j. j. , wegner g. a. , 2000 , , 120 , 2338 k. , geller m. j. , kurtz m. j. , diaferio a. , 2003 , , 126 , 2152 robertson , b. , yoshida , n. , springel , v. , & hernquist , l. 2004 , , 606 , 32 robertson , b. , hernquist , l. , bullock , j.s . , cox , t.j . , di matteo , t. , springel , v. , & yoshida , n. 2005 , , submitted salpeter , e. e. 1964 , , 140 , 796 sanders , d.b .
1986 , , 305 , l45 sanders , d.b .
1988a , , 325 , 74 sanders , d.b .
et al . 1988b , , 324 , l55 sanders , d.b .
et al . 1988c , , 328 , l35 sanders , d. b. & mirabel , i. f. 1996 , , 34 , 749 sargent , a.i .
1987 , , 312 , l35 sargent , a.i . , sanders , d.b . &
phillips , t.g .
1989 , , 346 , l9 sazonov , s. y. , ostriker , j. p. , ciotti , l. , & sunyaev , r. a. 2005 , , 358 , 168 schmidt , m. 1963 , nature , 197 , 1040 scoville , n.z .
1986 , , 311 , l47 silk , j. & rees , m. j. 1998 , , 331 , l1 soifer , b.t .
et al . 1984a , , 278 , l71 soifer , b.t .
et al . 1984b , , 283 , l1 soifer , b.t . ,
houck , j.r .
& neugebauer , g. 1987 , ara&a , 25 , 187 sokasian , a. , abel , t. , & hernquist , l. 2002 , , 332 , 601 sokasian , a. , abel , t. , & hernquist , l. 2003 , , 340 , 473 soltan , a. 1982 , , 200 , 115 springel , v. , di matteo , t. , & hernquist , l. 2005a , , submitted , [ astro - ph/0409436 ] springel , v. , di matteo , t. , & hernquist , l. 2005b , , submitted , [ astro - ph/0411108 ] springel , v. & hernquist , l. 2002 , , 333 , 649 springel , v. & hernquist , l. 2003 , , 339 , 289 springel , v. & hernquist , l. 2005 , , in press [ astro - ph/0411379 ] springel , v. , yoshida , n. , & white , s. d. m. 2001 , new astronomy , 6 , 79 stevens , j. a. , page , m. j. , ivison , r. j. , carrera , f. j. , mittaz , j. p. d. , smail , i. , & mchardy , i. m. 2005 , , submitted , [
astro - ph/0503618 ] stockton , a. 1978 , , 223 , 747 stockton , a. & mackenty , j. 1987 , , 316 , 584 stockton , a. & ridgway , s. 1991 , , 102 , 488 telfer , r. c. , zheng , w. , kriss , g. a. , & davidsen , a. f. 2002 , , 565 , 773 a. , 1977 , in evolution of galaxies and stellar populations , 401 , yale univ .
obs : new haven a. & toomre j. , 1972 , , 178 , 623 ueda , y. , akiyama , m. , ohta , k. , & miyaji , t. 2003 , , 598 , 886 vanden berk , d. e. , et al.2001 , , 122 , 549 vignali , c. , brandt , w. n. , & schneider , d. p. 2003
, , 125 , 433 webster , r. l. , francis , p. j. , peterson , b. a. , drinkwater , m. j. , @xmath162 masci , f. j. 1995 , , 375 , 469 white , r. l. , helfand , d. j. , becker , r. h. , gregg , m. d. , postman , m. , lauer , t. r. , @xmath162 oegerle , w. 2003 , , 126 , 706 white , s.d.m . & rees , m.j .
1978 , mnras , 183 , 341 wyithe , j. s. b. & loeb , a. 2002 , , 581 , 886 wyithe , j. s. b. & loeb , a. 2003 , , 595 , 614 yu , q. & lu , y. 2004 , , 602 , 603 yu , q. & lu , y. 2005 , , in press [ astro - ph/0411098 ] yu , q. & tremaine , s. 2002 , , 335 , 965 zeldovich , ya.b . & novikov , i.d . 1964 , sov . | based on numerical simulations of gas - rich galaxy mergers , we discuss a model in which quasar activity is tied to the self - regulated growth of supermassive black holes in galaxies . the nuclear inflow of gas attending a galaxy collision triggers a starburst and feeds black hole growth , but for most of the duration of the starburst , the black hole is `` buried '' , being heavily obscured by surrounding gas and dust , limiting the visibility of the quasar , especially at optical and ultraviolet wavelengths . as the black hole grows , feedback energy from accretion heats the gas and eventually expels it in a powerful wind , leaving behind a `` dead quasar '' . in between the buried and dead phases
, there is a window in time during which the galaxy would be seen as a luminous quasar .
because the black hole mass , radiative output , and distribution of obscuring gas and dust all evolve strongly with time , the duration of this phase of observable quasar activity depends on both the waveband and imposed luminosity threshold .
we determine the observed and intrinsic lifetimes as a function of luminosity and frequency , and calculate observable lifetimes @xmath0myr for bright quasars in the optical b - band , in good agreement with empirical estimates and much smaller than our estimated black hole growth timescales @xmath1myr , naturally producing a substantial population of buried quasars .
however , the observed and intrinsic energy outputs converge in the ir and hard x - ray bands as attenuation becomes weaker and chances of observation greatly increase .
we also obtain the distribution of column densities along sightlines in which the quasar is seen above a given luminosity , and find that our result agrees remarkably well with observed estimates of the column density distribution from the sdss for the appropriate luminosity thresholds .
our model reproduces a wide range of quasar phenomena , including observed quasar lifetimes , intrinsic lifetimes , column density distributions , and differences between optical and x - ray samples , having properties consistent with observations across more than five orders of magnitude in bolometric luminosity from @xmath2 ( @xmath3 ) . |
interferometers , which are among the most precise measurement devices known today , are based on the interference of two coherent atomic or photonic modes .
the quantity to be measured is mapped to the relative phase between the two modes , explicitly using the wave - like nature of the resources ( atoms or photons ) . however , in the actual detection process their particle - like character becomes important ; the population of the two modes can be at best be counted one by one .
traditional interferometers use uncorrelated resources in two coherent modes which results in a poissonian distribution of the population difference at the interferometer output .
the standard deviation of this distribution @xmath0 is the so called projection noise or shot noise limit .
the best theoretically possible phase precision for uncorrelated resources and linear interferometers is given by the classical cramer - rao bound of parameter estimation @xcite .
in the context of ( classical ) interferometric phase estimation this is known as the standard quantum limit .
it can be reached with coherent states , the quantum states that are naturally generated by coherent splitting of a quantum mode , as it is routinely done by beam splitters in optics ( with one input port unused ) .
indeed , modern interferometers measuring for example time @xcite , position @xcite or magnetic fields @xcite are limited by this quantum noise leading to a maximal interferometric phase precision of @xmath1 .
several strategies have been proposed to overcome this standard quantum limit for metrology by using quantum correlated , that is entangled , resources within the interferometer @xcite .
one of these strategies is to use squeezed quantum states @xcite which have been successfully generated in quantum optics @xcite .
the analogue for atom interferometry are spin squeezed states @xcite .
spin squeezed states can be obtained from coherent spin states by a continuous transformation which gradually redistributes the quantum noise among two orthogonal spin directions .
this property makes them promising candidates for their experimental realization since the degree of squeezing can be balanced with the increased sensitivity to decoherence .
interferometric precision is increased to @xmath2 for spin squeezed states with a squeezing parameter @xmath3 .
many - body quantum mechanics sets an ultimate limit to the phase precision for general interferometric measurements .
this is the so - called heisenberg limit , for which an @xmath4 scaling of the interferometric precision can be achieved @xcite .
the heisenberg limit can be reached with spin squeezed states in the limit of infinite squeezing where spin squeezed states become identical to twin fock states .
the potential gain in phase precision is enormous : in an atomic clock , for example , one measures an energy difference @xmath5 for a certain time @xmath6 ( @xmath7 is planck s constant ) .
the `` classical '' uncertainty scales as @xmath8 , while heisenberg limited metrology would allow for an error of @xmath9 . assuming a fictitious heisenberg limited measurement with @xmath10 atoms lasting one second
, a `` classical '' projection noise limited apparatus would need @xmath11 days to obtain the same level of precision .
+ in this tutorial we attempt to provide a basic introduction into the concept of spin squeezing .
we focus on the theoretical concept of spin squeezing , its connection to entanglement and applications in quantum metrology . experimental techniques to generate and detect spin squeezing in bose - einstein condensates ( bec ) are explained in detail .
finally we shortly discuss two exemplary experiments in which spin squeezing has been observed in becs .
spin squeezing is a quantum strategy introduced in 1993 by kitagawa and ueda @xcite which aims to redistribute the fluctuations of two orthogonal spin directions among each other . in 1994
it was theoretically shown that spin squeezed states are useful quantum resources to enhance the precision of atom interferometers @xcite and in 2001 the connection between spin squeezing and entanglement was pointed out @xcite . in the field of quantum optics
squeezed states have been proposed to overcome the standard quantum limit in optical interferometry already in 1981 @xcite .
these states are similar to spin squeezed states however differences arise from the different variables , quadratures versus collective spin directions @xcite .
the connection of spins to interferometry becomes obvious when realizing that an interferometer is essentially a two mode system with modes @xmath12 and @xmath13 .
each particle in the interferometer can be regarded as an elementary spin with spin @xmath14 , with two possible longitudinal eigenstates @xmath15 corresponding to the two interferometer modes .
the relative phase measured by the interferometer corresponds to the spin component in the plane spanned by the transversal spin directions @xmath16 and @xmath17 .
the generalization of this picture to @xmath18 particles , which results in large collective spin vectors @xmath19 , is outlined below .
an extensive theoretical review of two - mode systems also detailing the pseudo spin description can be found in @xcite .
the mathematical concept of a spin algebra with total spin @xmath20 is a powerful tool to describe very different physical systems .
any observable within such a spin @xmath20 system can be expressed by the three spin operators @xmath21 , @xmath22 , @xmath23 and the identity operator .
the collective spin operators @xmath24 can be defined as the sum over all elementary spin operators ( pauli matrices ) @xmath25 , where @xmath26 : @xmath27 the three orthogonal spin components are non - commuting operators .
their commutation relation is @xmath28=i \,\epsilon_{ijk } { \ensuremath{\hat{\textsf{\ensuremath{j}}}}}_{k}$ ] , where @xmath29 is the levi - civita symbol .
therefore , any pair of spin operators obeys a heisenberg uncertainty relation which , for @xmath30 and @xmath31 , is given by @xmath32 and @xmath33 is the variance in @xmath34 direction .
a basis of the general problem can be obtained as the tensor product of all @xmath18 bases of the individual components , each of dimension @xmath35 the dimension of the hilbert space is huge @xmath36 and grows exponentially with the number of elementary spins .
the length of the collective spin @xmath20 is smaller or equal than half the number of elementary spins .
the spin length in interferometry is usually large such that @xmath37 is a good approximation . in the symmetric case ,
that is , when all operations done on the spin ensemble affect each spin in the same way ( such that there is exchange symmetry ) the problem simplifies considerably .
for example , each elementary spin can be prepared in the @xmath38 state and maximum collective polarization @xmath39 can be reached .
hence , the total spin length is given by @xmath40 and the dimension of the hilbert space dramatically reduces to @xmath41 , linearly dependent on the number of elementary spins .
one possible choice of a basis are the symmetric dicke states @xcite @xmath42 with @xmath43 .
each of the dicke states corresponds to a perfectly defined particle number difference between the two modes @xmath12 and @xmath13 and since the total number of particles @xmath18 is fixed
the dicke states correspond to twin fock states in the modes @xmath12 and @xmath13 .
due to their exchange symmetry the elementary spins can be effectively described as bosons , the schwinger bosons @xcite . employing the second quantization formalism the creation ( @xmath44 , @xmath45 ) and annihilation operators ( @xmath46 , @xmath47 ) of the two modes
can be related to the different spin components @xcite @xmath48 with @xmath49 and @xmath50 .
identical bosonic particles in two modes can be described by the symmetric spin model and the schwinger representation given above can be used . when it comes to experiments
, however , an a priori symmetry assumption is a strong constraint .
its validity has to be shown experimentally .
coherent spin states are the most `` classical - like '' pure quantum states of a symmetric ensemble of @xmath18 elementary @xmath51 spins or of @xmath18 two - mode bosons @xcite .
they are constructed by placing all @xmath18 particles in the same single particle state , in any superposition of the two modes @xmath52^{n } { \ensuremath{\left| \mathrm{vac } \right\rangle}}\ ] ] where @xmath53 is the vacuum state .
any coherent spin state @xmath54 can be constructed from the state @xmath55 ( all n bosons in mode @xmath13 ) by a suitable unitary rotation with the angles @xmath56 @xcite .
one important aspect of the coherent spin states is that no quantum correlations between the particles are present .
therefore , a coherent spin state features equal variance in any direction @xmath57 orthogonal to the mean spin direction @xmath56 which , according to the gaussian addition of variances , is given by the sum of the variances of the @xmath58 elementary spin @xmath59 particles .
the perpendicular variances @xmath60 of individual elementary spins are by definition isotropic around @xmath56 since there are no subsystems that could cause any correlations ( figure [ fig : css]a ) @xcite .
the heisenberg uncertainty relation ( [ eq.spinheisenberg ] ) for a single elementary spin @xmath61 pointing in @xmath62 direction is @xmath63 leading to an isotropic variance of @xmath64 for the collective coherent spin state , which identifies this quantum state as a minimal uncertainty state since @xmath65 ( see equation ( [ eq.spinheisenberg ] ) ) .
we refer to the perpendicular spin fluctuations of a coherent spin state @xmath66 as the shot noise limit . + we go back to the first quantization formalism in order to obtain the probability distribution over different sets of basis states especially the two dicke state bases in the directions orthogonal to the mean spin direction . in order to develop a more detailed understanding of the coherent spin state and its fluctuations we start with the discussion of a special case where each particle is in a @xmath67 superposition of the two modes with @xmath68 relative phase .
each elementary spin points in @xmath62 direction and its quantum state is @xmath69 , where we have chosen the dicke states in @xmath70 direction as the basis states .
the probability to observe the elementary spin in state up or down is equal @xmath71 . +
the @xmath18 atom coherent spin state is a collection of these independent elementary spins @xmath72^{\otimes n}$ ] and hence the measurement of the @xmath73 spin component is equivalent to @xmath18 measurements on a single spin .
the probability distribution over the dicke states is binomial .
we could have chosen equally the dicke states in @xmath74 direction to describe the spin state which shows again that the spin fluctuations in the directions perpendicular to @xmath75 , the mean spin direction , are isotropic . + a general coherent spin state @xmath76 in the dicke basis @xmath42 is described by @xcite : @xmath77 the coefficients @xmath78 follow a binomial distribution peaked around @xmath79 : .
+ [ t ] due to the fundamental uncertainty in the orthogonal spin directions the mean direction ( @xmath80 ) of any spin state can not be defined with arbitrary precision . for a coherent spin state the isotropic angular uncertainty @xmath81 , defined by the ratio of the uncertainty of the perpendicular spin directions
@xmath82 to the mean spin length @xmath20 , is given by : @xmath83 as outlined above this limit arises as the classical statistical limit in a system consisting of @xmath18 independent constituents @xcite .
it is this angular uncertainty which limits the interferometric precision in a ramsey interferometer .
quantum correlations between the elementary spin @xmath59 particles of a collective spin @xmath20 can cause anisotropic fluctuations of the collective spin in the directions perpendicular to the mean spin ( figure [ fig : css]b ) . in reference @xcite quantum states
are considered spin squeezed if the variance of one spin component is smaller than the shot noise limit @xmath84 for a coherent spin state : @xmath85 definition ( [ eq.numbersqueezing ] ) does not take the second , perpendicular spin direction into account .
due to the heisenberg uncertainty relation ( [ eq.spinheisenberg ] ) reduction of the variance in one direction causes an increase of fluctuations in the other . real life strategies to obtain spin squeezing might also involve states that are not minimal uncertainty states . for these states , especially for experimentally very important non - pure quantum states ,
the variance in some other direction than the squeezed direction can be much larger than given by the heisenberg uncertainty relation .
this leads to a reduction of the effective mean spin length @xmath86 .
+ standard metrologic applications , for example ramsey interferometry , require a large mean spin length . in order to measure the usefulness of spin squeezed states for these applications another definition of the squeezing parameter
was introduced in reference @xcite @xmath87 whose inverse @xmath88 measures the precision gain in a ramsey interferometric sequence relative to the standard quantum limit ( [ eq.shotnoise ] ) .
spin squeezing among @xmath18 constituents is related to many - body entanglement . in this context
a third spin squeezing criterion was defined @xcite : @xmath89 entanglement is detected by the inequality @xmath90 as detailed below .
here we explicitly use the standard assumption throughout this tutorial that the mean spin points in @xmath75 direction and the direction of minimal variance if not explicitly mentioned is the @xmath73 direction .
+ @xmath91 can be used equivalently to @xmath92 to quantify spin squeezing and precision gain in interferometry and we refer to it as coherent number squeezing or coherent spin squeezing .
both quantities remain a useful parameter to detect entanglement or to measure metrologic precision gain even in a non - symmetric situation . to demonstrate entanglement in a collective spin system without presumptions not only number squeezing is required but also the coherence of the quantum state has to be shown . employing the q - representation @xcite ,
collective spin states can be conveniently visualized on a generalized bloch sphere with radius @xmath20 . in order to describe the most general spin state , i.e. pure states and statistical mixtures ,
the density matrix formalism is used @xcite . the density operator @xmath93 in coherent spin state basis is given by @xmath94 , where the integral covers the full solid angle and @xmath95 .
the quasi - probability distribution @xmath96 is normalized to one .
the q - representation uses the diagonal elements of the density operator to represent the quantum state @xmath97 .
it gives the probability of finding the system specified by @xmath93 in the coherent spin state characterized by @xmath98 .
the interpretation of this representation on generalized bloch spheres differs from the usual single spin @xmath51 bloch sphere . in that case
the dimension of the hilbert space is two - dimensional and the quantum state representation on the surface of a sphere is exact .
however for symmetric collective spin systems the dimension of the hilbert space is @xmath99 such that an exact mapping to the surface of a sphere is not possible . the position @xmath100 on the spin @xmath59 bloch sphere describes the full quantum state , while the position on the generalized bloch sphere gives only the mean spin direction and , within the following constraints , its fluctuations : the q - representation projects the density matrix on minimal uncertainty states , in particular coherent spin states .
the most obvious consequence is that the minimal extension of a quantum state in @xmath101 on the bloch sphere is given by the uncertainties of the basis states .
+ the q - representation for visualizing collective spins is rather easily implemented with modern plotting tools .
however , there exist more elaborate ways to visualize the quantum state of the two - mode system as described in @xcite .
the bloch vector visualization described therein has the advantage that the spin noise in all three components of the bloch vector is directly visible .
also the fragmentation of the quantum state , that is , if the @xmath18 bosons do not occupy the same single particle state @xcite can be seen directly by a bloch vector that lies inside the bloch sphere .
entanglement is only a meaningful concept if the subsystems relative to which entanglement is defined are clearly specified @xcite .
coherent spin squeezing detects entanglement among the elementary spins the subsystems are the individual spin @xmath59 particles @xcite . for @xmath18 distinguishable particles the definition of a separable state , i.e. non - entangled state , is that its @xmath18-body density matrix @xmath102 can be written as a direct product of single particle density matrices @xmath103 : @xmath104 @xmath105 is a probability distribution to account for incoherent mixtures .
entanglement in many - body systems ( for a general review see @xcite ) is defined as the non - separability of the density matrix @xmath102 meaning the equality in equation ( [ eq.separability ] ) does not hold .
+ due to technical limitations the individual elementary spins can not be addressed before the measurement in many experiments .
however , it is important to note that the elementary spins ( the subsystems ) have to be in principle distinguishable in order to define entanglement among them in a meaningful way @xcite .
for example , for @xmath18 particles in a bec the distinguishability is a priory not given .
however , by a gedanken local operation one can pinpoint each particle in space at the time of measurement without affecting the spin properties of the system .
distinguishability is now given by the position of each particle .
if entanglement is detected in the system , it must have been present in the system before the localization , since local measurements can not generate entanglement @xcite . given that the atoms in the bec are spaced by more than one wavelength of the detection light ( which is usually fulfilled ) , this gedanken local operation means to overcome the technical limitations for addressability and detection of the individual elementary spins .
the fictitious pinning of the particles at the time of measurement ( the photon scattering destroys the bec ) effectively assigns a spatial mode to each particle by which they become distinguishable @xcite .
the authors of reference @xcite show that the notion of entanglement in such a collective spin system can be generalized even without such a fictitious pinning of the individual particles .
+ not surprisingly different entanglement criteria can be found when regarding the two modes that define the spin @xmath59 , as the subsystems @xcite .
it has been argued that this kind of mode entanglement can not be extracted in experiments without a suitable reference frame @xcite .
recently , this problem of a missing reference frame for atomic fields has been addressed experimentally and mode entanglement has been shown in a system of ultracold bosons @xcite . without the possibility to address the individual elementary spins before the measurement tests for entanglement ,
that is , experimentally accessible criteria , based on the collective spin variables are necessary .
the observables in most experiments so far are limited to first and second order moments of the distributions functions in different spin directions due to rather small counting statistics and technical noise .
based on these , a complete set of inequalities , that is fulfilled for any separable quantum state , has been found @xcite .
complete in this sense means that assuming the only information available are first @xmath106 and second moments @xmath107 of the distribution functions .
note that these inequalities allow for experimental tests of entanglement , but they are no quantitative measures for entanglement .
such entanglement criteria only allow for the statement : `` yes , that state is entangled with respect to these subsystems . '' or
`` no , that state is not entangled with respect to these subsystems . ''
the coherent spin squeezing criterion given in equation ( [ eq.coherentspinsqueezing ] ) corresponds to a less general form of such an inequality : all separable states fulfill the inequality @xmath108 , but a subgroup of entangled states violate it @xcite . coherent spin squeezing as an entanglement criterion does not require any symmetry assumption .
however , in the fully symmetric subspace many entanglement criteria simplify .
for example , the detection of spin fluctuations in one direction below the shot noise limit implies entanglement @xcite , that is @xmath109 holds for any separable symmetric state .
equation ( [ eq.sqsymmetric ] ) is identical to the spin squeezing parameter of kitagawa and ueda ( [ eq.numbersqueezing ] ) showing that in the symmetric two - mode case entanglement is necessary to redistribute the fluctuations of orthogonal spin components . to avoid misunderstandings about the different definitions of spin squeezing
we refer to @xmath110 as number squeezing .
entanglement in collective spin systems is not only interesting from a conceptual perspective but it has also been shown to provide a useful quantum resource . in 1994
wineland et al .
@xcite pointed out , that in particular spin squeezed states can be used to overcome the standard quantum limit in metrology .
they can be directly used to boost the precision of conventional ramsey interferometers .
prominent applications of ramsey interferometry @xcite are the definition of the time standard @xcite , high precision magnetometry @xcite or inertia measurements @xcite .
reviews covering atom interferometry both experimentally and theoretically can be found in references @xcite . + figure [ fig : ramsey_bs ] shows the ramsey interferometric sequence illustrated on generalized bloch spheres .
the parameter @xmath111 to be measured couples via a @xmath23 term in the hamiltonian resulting in a rotation of the spin around the @xmath34 axis during the evolution with rotation frequency @xmath111 . the accumulated relative phase @xmath112 by which the initial spin state rotated around the @xmath34 direction of the bloch sphere is finally converted into an observable population difference .
the phase resolution of the interferometer @xmath113 is determined by the projection noise @xmath114 at the time of readout and the local slope of the ramsey fringe @xmath115 ( see figure [ fig : ramsey_prec ] ) .
the point of maximum sensitivity ( taking finite environmental noise into account ) is reached where the mean population difference is zero and the slope is maximal @xmath116 . the visibility @xmath117 measures the mean spin length @xmath118 . + the amount of precision gain ( or loss ) relative to standard quantum limit is given by @xmath119 , or equivalently by @xmath120 .
the measure can be expressed in visibility @xmath117 and spin noise in @xmath73 direction at readout @xmath30 @xmath121 the phase uncertainty measured as the root mean square deviation is @xmath122 .
+ the phase @xmath123 appears in two connected contexts in this tutorial . on the one hand , for coherent spin states , it is a variable characterizing the full quantum state of the system ( together with @xmath79 ) . on the other hand , it has the meaning of a relative phase between the two modes of the interferometer ( or more general between the modes @xmath12 and @xmath13 of the two - mode system ) . in both cases
the phase @xmath123 can be interpreted as the mean angle of the spin state in longitudinal direction ( that is , in the @xmath124 , @xmath125 plane ) on the bloch sphere as schematically shown in figure [ fig : ramsey_bs ] .
the angular uncertainty in this direction finally limiting the precision of a ramsey interferometers is @xmath126 ( cf .
section [ sec.css ] ) .
spin squeezed states feature reduced noise in one of the spin directions but excess noise in another direction can be present either due to a non - heisenberg limited quantum state or due to an incoherent mixture of several quantum states .
the former might limit precision in standard ramsey interferometry , but specific correlated quantum states enable even enhanced interferometric precision in a generalized interferometer @xcite .
the latter is easily limiting interferometric precision at a level above the standard quantum limit and experimentally it requires a large effort to prevent decoherence due to technical noise from the environment or due to finite temperature in the system .
large noise quantum or classical even in a spin direction that is not directly measured has a degrading effect on interferometric precision in a standard interferometer which arises due to the curved surface of the bloch sphere .
as soon as the noise amplitude is large enough such that the area of uncertainty can no longer be approximated by a plane , the mean spin is effectively shortened and the visibility decreases @xmath127 .
+ entanglement can be used as a quantum resource in a ramsey interferometric sequence in different ways . in order to increase the phase sensitivity either the slope of the signal
@xmath128 has to be increased or the projection noise @xmath30 has to be decreased .
slope increase can be reached by schrdinger cat type entanglement which involves maximally entangled states that are very fragile to decoherence .
therefore they have been realized so far with very few particles only @xcite .
+ spin squeezing aims to decrease the projection noise .
this is possible in gradual steps meaning that depending on the amount of spin squeezing the precision is gradually increased .
therefore at least for moderate levels of spin squeezing these states are less fragile and they have been realized with a large number of particles but only with a relatively small squeezing factor @xcite .
interferometric sensitivity for a coherent spin state and a coherent spin squeezed state is illustrated in figure [ fig : ramsey_prec ] .
for the coherent spin squeezed state the decreased quantum fluctuations @xmath114 reduce the projection noise while the increased fluctuations @xmath129 cause a slight decrease of the mean spin length and therefore of the visibility of the ramsey fringe . nevertheless , the ratio of projection noise and slope of the ramsey fringe and therefore the phase sensitivity is increased .
the ultimate limit for metrologic precision is the heisenberg limit @xcite , where the phase estimation error @xmath130 is given by @xmath131 for @xmath18 resources used in a single measurement .
this fundamental limit can in principle be reached with both approaches mentioned above , schrdinger cat type entanglement or spin squeezing . in the context of quantum metrology the schrdinger cat state
is sometimes called a noon state @xcite .
its name originates from its form in fock states basis @xmath132 .
it is a coherent superposition of all atoms in state @xmath12 and zero atoms in state @xmath13 and vice versa . in spin
representation the noon state is the superposition of the two maximal dicke states @xmath133 .
the increase of the signal slope for a noon state is obvious since the phase acquired between the two components @xmath134 is @xmath18 times larger than for a coherent spin state @xcite .
experimentally it is important to note that the readout of the interferometer can not be realized by measuring @xmath135 .
the reason is the vanishing mean spin length @xmath136 of this state .
it has been shown that the parity of the state is a useful experimental observable to make use of noon states in interferometry and to reach the heisenberg limit @xcite .
further proposals to achieve heisenberg limited interferometry involve twin - fock states @xmath137 @xcite .
the mean spin length for these states is also vanishing such that special readout schemes are required @xcite .
recently such states have been produced experimentally @xcite and their usefulness for quantum enhanced metrology has been shown @xcite .
spin squeezed states allow for the optimization of interferometry gain demanding a finite mean spin length such that standard interferometric readout can be used .
the optimum @xmath92 for a given mean spin length was found numerically in reference @xcite .
an experimental protocol to generate spin squeezed states close to the heisenberg limit was proposed in reference @xcite .
the authors show that the heisenberg limit can be reached asymptotically in the number of atoms .
the precision of ramsey type interferometric measurements increases with evolution time within the interferometer ( cf .
section [ sec.intro ] ) .
the same inter atomic interactions , that can be used to generate entangled states initially and to overcome the standard quantum limit ( as shown in this tutorial ) , however , might spoil the precision of the interferometer if present during this evolution time @xcite .
hence , interaction control ( for example using feshbach resonances ) is crucial for future applications that use standard interferometric readout .
recently it has been emphasized that the fisher information is the most general criterion to measure phase sensitivity @xcite since it saturates the quantum cramer - rao bound @xcite . calculating the fisher information for a coherent spin state state evolving under the non - linear hamiltonian @xmath138 , where @xmath139 parametrizes the nonlinearity , heisenberg like scaling for the phase precision was found @xcite after long evolutions times where the potential precision gain in a standard interferometer was already decreased .
the quantum state here is neither necessarily a noon state nor a coherently spin squeezed state .
however standard interferometric readout can not be used to extract the phase information and a new type of bayesian readout might be employed which was experimentally demonstrated in @xcite .
strong effort has been put into the experimental generation and detection of spin squeezed states in large collective spin systems .
early experiments detected spin noise beyond the shot noise limit in a cold ensemble of cesium atoms prepared by quantum state transfer from non - classical light to the atoms @xcite , or in a cesium vapor cell at room temperature by quantum non - demolition measurement @xcite .
indirect measurements of number squeezing of ultracold atoms in optical lattices and double well systems followed , however , the direct detection of the fluctuations was not possible @xcite . several experiments generating and detecting coherent spin squeezing of large collective spins in atomic gases were done in last four years @xcite .
full ramsey interferometers working beyond the standard quantum limit have been realized using coherent spin squeezed states within the interferometer @xcite .
different methods have proven useful to generate spin squeezed states in atomic samples .
many of them involve light , for example by quantum state transfer , quantum non - demolition measurements or cavity induced interactions @xcite . in contrast , ultracold quantum gases offer the possibility to directly use the intrinsic interaction between the atoms .
these interactions lead to nonlinear terms in the hamiltonian necessary to generate spin squeezed states @xcite .
we will take intrinsic s - wave interactions among atoms in a two - mode bec as an example here and discuss how spin squeezing can be generated .
examples for such two - mode systems based on internal or external degrees of freedom are shown in figure [ fig : example_twomode ] .
in such two - mode becs the inter atomic interactions lead to a nonlinear term in the hamiltonian proportional to @xmath140 @xcite .
a review on the experimental realization of the external two mode system can be found in @xcite and for the internal system details are given in @xcite .
the theoretical description for both cases , including the possibility for fragmentation of the quantum state @xcite , has been reviewed in detail recently @xcite .
thus , we only sketch the connection between the general hamiltonian and the pseudo spin description for the external double well case in the following
. the general hamiltonian is @xmath141 from which , by making the two - mode ansatz @xmath142 , the two - mode bose - hubbard hamiltonian is obtained ( for improved readability we did not write the spatial dependence of the operators in equation ( [ eq.genham ] ) explicitly ) .
@xmath143 here we defined three parameters in the hamiltonian , the nonlinearity @xmath139 characterizing the inter atomic interactions , the rabi type coupling between the two modes @xmath144 and a differential energy shift @xmath111 .
the remaining step is to write the hamiltonian ( [ eq.bosehubbard ] ) in spin language by using the schwinger representation introduced in equation ( [ eq.schwinger ] ) : @xmath145 [ t ] and @xmath146 define the modes . * b * another example for a two - mode bec is a spinor bec with two populated hyperfine spin states @xmath12 and @xmath13 and identical spatial wave function @xmath147 . ]
note that the hamiltonian in both cases , external and internal degrees of freedom , can be written in the form of equation ( [ eq.spin ] ) .
the actual calculation of the parameters involves spatial integrals over the wave functions and differs in both cases . in this context
it is important to note that the spatial wave functions @xmath147 are not the single particle ground state wave functions in the confining potential , but the mean field wave functions obtained by coupled gross - pitaevskii equations for the two modes ( for several different approaches see @xcite ) .
details on the calculation of the parameters in the hamiltonian can be found for example in @xcite .
+ three regimes can be distinguished by looking at the quantum fluctuations in the orthogonal spin directions for the ground state of the hamiltonian in equation ( [ eq.spin ] ) ( see figure [ fig : adiabatic_scheme ] ) . for the following discussion
we assume an unbiased situation , that is @xmath148 .
the important parameter measuring the relative importance of tunnel coupling and interaction is @xmath149 .
this can be seen directly by normalizing the spin operators to spin @xmath59 such that the total number of atoms appears explicitly .
then @xmath150 is the ratio of the nonlinear to the linear coefficient .
the rabi regime is the noninteracting limit where @xmath151 . here
the ground state is a coherent spin state with maximal mean spin length @xmath152 and equal spin fluctuations in the orthogonal directions @xmath153 . for increasing interactions fluctuations in the relative atom number ( equivalent to fluctuations in @xmath23 )
become energetically unfavorable and the system enters the josephson regime ( @xmath154 ) .
its name originates from the analogy of the hamiltonian ( [ eq.spin ] ) to the josephson hamiltonian in solid state physics @xcite .
the ground state is a coherent spin squeezed state with reduced spin fluctuations in @xmath34 direction at the cost of increased fluctuations in @xmath125 and hence a reduced mean spin length .
the tunnel coupling is negligible in the fock regime ( @xmath155 ) .
the ground state is an extremely spin squeezed state with vanishing mean spin length .
it finally approaches the twin fock state , an eigenstate of @xmath23 with eigenvalue zero .
there are two possibilities to control the regime parameter @xmath150 . for a given total atom number @xmath18 either the tunnel coupling @xmath144 or the nonlinearity
@xmath139 has to be changed . while the former is straight forward , there are again several possibilities to change the nonlinearity @xmath139 .
the rather direct approach is to control the scattering properties @xcite , which is directly possible when using quantum states with accessible feshbach resonances @xcite .
the second method , which has also been realized experimentally , is to control the wave function overlap between the two states @xcite . assuming the weakly interacting regime
can be realized experimentally there are two strategies to generate coherent spin squeezed states .
the first is an adiabatic approach where the regime parameter is changed slowly driving the system deeper into the josephson regime @xcite .
for some experimental settings a fast non - adiabatic approach might be preferable @xcite where spin squeezing develops during the evolution after a quench in @xmath150 . in
the following we discuss the advantages and disadvantages of the two strategies .
[ t ] one of the advantages of ultracold atom systems is the ability to control the parameters of the hamiltonian dynamically .
starting from a strong tunnel coupling the regime parameter @xmath150 can be adiabatically changed .
such an adiabatic transformation of the quantum state is a robust way to generate coherent spin squeezed states .
the tunnel coupling is present throughout the transformation , making the system less prone to phase noise ( fluctuations in @xmath111 ) .
this can be intuitively understood by the concept of the spin echo technique @xcite : for small differential energy shifts @xmath156 the continuous tunnel coupling causes a continuous sequence of @xmath157 echo pulses counteracting the dephasing .
another advantage of the adiabatic technique is its usefulness in a finite temperature environment , where the initial state is a thermal mixture of the lowest many - body eigenstates of the hamiltonian ( [ eq.spin ] ) . in the josephson regime
the adiabatic increase of @xmath150 results in an effective `` cooling '' of the system @xcite .
the term cooling should be used with care here since no entropy is removed from the system . with increasing parameter @xmath150 each of the lower eigenstates of the system
features reduced fluctuations in @xmath23 , however at the cost of increased fluctuations in @xmath22 and a loss in coherence .
nevertheless , the gain in @xmath34 direction outweighs the loss of coherence for small enough temperatures , such that the coherent spin squeezed regime can be reached by this method .
it has been theoretically shown that the heisenberg limit can be reached asymptotically in the number of particles by the adiabatic technique @xcite . experimentally , however , it is a major problem to assure adiabaticity during the full range of the parameter change .
at least when changing the regime parameter by tuning the tunneling coupling ( for fixed and typically small nonlinearity @xmath139 ) the required timescale for adiabaticity eventually diverges setting a limit to the attainable coherent spin squeezing factor @xmath3 .
the long times required to generate coherent spin squeezing adiabatically are the main disadvantages of this approach . for some experimental systems
the robust but slow adiabatic approach to generate spin squeezing might not be optimal . a diabatic technique , in which the hamiltonian parameters are quenched ( changed non - adiabatically ) , can produce spin squeezing in a much shorter time .
the idea of the diabatic one - axis twisting was introduced in 1993 by kitagawa and ueda . starting with a coherent spin state with mean spin in @xmath124 direction the tunnel coupling
is suddenly switched off such that the nonlinear term in the hamiltonian governs the evolution @xmath158
the nonlinearity causes a shearing of the uncertainty region on the bloch sphere since the rotation speed around @xmath34 is itself @xmath34 dependent .
the scheme is similar to kerr effect based squeezing protocols in quantum optics where a material with a finite kerr nonlinearity is used such that the refractive index @xmath159 is proportional to the light intensity @xmath160 .
the light experiences intensity dependent phase modulation within this medium resulting in quadrature squeezing @xcite .
[ t ] the one axis twisting evolution of an initial coherent spin state can be nicely visualized . earlier in this manuscript
it has been discussed how a coherent spin state on the equator of the bloch sphere @xmath161 can be described as a coherent superposition of several dicke states where the probability distribution over these basis states is binomial .
each dicke state @xmath42 composing the coherent spin state rotates with a different frequency around the @xmath73 axis where the difference in rotation frequency between next neighboring dicke states is @xmath139 ( figure [ fig : diabatic_scheme]a ) . for short evolution times
@xmath162 this shearing effect results in coherent spin squeezing under an axis rotated by the angle @xmath163 with respect to the equator of the bloch sphere ( figure [ fig : diabatic_scheme]b ) .
@xmath164 is the dephasing time , after which the mean spin length has dropped from a value close to unity to @xmath165 due to the interaction induced spread of the state around the bloch sphere @xcite . in general @xmath166
is inversely proportional to the extension of the quantum state over the dicke basis @xmath167 which , for a coherent spin state on the equator of the bloch sphere , is @xmath168 .
+ at evolution times @xmath169 reduced quantum fluctuations under a certain axis are still present but the mean spin length is strongly reduced such that the quantum state is no longer optimally coherently spin squeezed .
after even longer times the dynamics show highly non - classical interference effects @xcite finally resulting in a revival of the mean spin length at @xmath170 when each neighboring pair of dicke states is in phase again @xcite .
the best achievable noise suppression in general interferometry still increases with evolution time even if coherent spin squeezing degrades .
it has been recently shown @xcite that a new type of bayesian interferometer readout can be employed to make use of these quantum states . as typical for macroscopic superposition states the highly entangled states occurring for long evolution times are extremely sensitive to decoherence .
for example , particle loss leads to an exponential decay of the revival contrast with a @xmath18 times enhanced decay constant @xcite .
the maximum squeezing achievable with the diabatic one axis twisting technique is @xmath171 @xcite .
this has to be contrasted to the maximum squeezing that can be generated using the adiabatic technique which approaches the heisenberg limit @xmath172 at the boundary to the fock regime where the mean spin length is still reasonable high @xcite .
the maximum squeezing is better in the adiabatic case however the adiabaticity criterion requires long evolution times @xmath173 .
the best squeezing in the one axis twisting protocol is achieved after @xmath174 , a factor of @xmath175 faster favoring the diabatic protocol for certain applications .
the generation of spin squeezing might be optimized using different hamiltonians .
stronger maximal squeezing , for example , can be reached by a two axis twisting scheme @xcite .
another example , optimizing the required time to achieve squeezing , is to use evolution under the hamiltonian ( [ eq.spin ] ) after a quench to a small but finite tunnel coupling ( into the josephson regime ) @xcite . for initial states with relative phase
@xmath176 the evolution leads first to coherent spin squeezing and for longer times to schrdinger cat like states @xcite .
when discussing the maximum achievable coherent spin squeezing one has to keep in mind that with decreasing @xmath177 the generated states become more and more sensitive to decoherence . for the exemplary case of the one axis twisting evolution
this problem has been studied in detail taking into account particle loss , random dephasing ( for example due to noise in @xmath111 , see equation ( [ eq.spin ] ) ) and multi mode effects due to finite temperature which are always present in real condensates @xcite . including these effects the achievable coherent spin squeezing is no longer proportional to @xmath178 but tends to a finite but small constant value in the thermodynamic limit @xmath179 , showing that a large metrology gain can still be reached @xcite .
reference @xcite gives a review on decoherence effects in dynamically generated spin squeezing with becs and points out the effective dephasing effect of the different sources of decoherence .
both approaches the adiabatic and diabatic one outlined in the last section require precise control over the experimental parameters .
especially so far neglected uncontrolled differential energy shifts @xmath111 are the main experimental challenge to observe spin squeezing .
another important issue , mainly in experiments with becs in double well traps , is finite temperature which easily limits the observable spin squeezing .
the detection of atomic noise at the shot noise level in a small optically thick cloud is challenging and requires careful calibration of the imaging system .
furthermore , a method to detect the mean spin length has to be available experimentally in order to detect coherent spin squeezing . in this section
we will discuss these challenges in detail , starting with the detection issues .
when working with internal degrees of freedom the measurement of the mean spin length can be done by a ramsey experiment @xcite .
a ramsey scheme is usually relatively easy to implement given the availability of precise microwave and radio frequency sources ( or raman lasers ) to control the tunnel coupling @xcite .
this coupling @xmath144 usually can be made much stronger than the nonlinearity @xmath139 such that the rabi regime is reached .
when working with two photon transitions , which is often required to use states less sensitive to magnetic field fluctuations , care has to be taken to accurately determine the resonance . the resonance is shifted during the coupling with respect to the free evolution times due to the ac - stark shift induced by the coupling fields @xcite .
the readout of the ramsey interferometer requires precise detection of the atom number difference in the two internal states over the full dynamic range .
the mean spin length @xmath180 is then inferred from the visibility @xmath117 of the ramsey signal .
an example for such a ramsey fringe revealing a visibility of @xmath181 is shown in figure [ fig_ramseyfringe ] @xcite .
[ t ] for external degrees of freedom , for example in a double well trap for becs the mean spin length can be revealed from the interference pattern after time of flight expansion @xcite .
additional complications might arise from a finite imaging resolution or due to excitations of additional spatial modes when expanding the condensate , both of which limit the visibility of the interference pattern in a single shot .
care has to be taken when extracting the average visibility @xmath117 . in general the analysis method chosen
has to include the effects of the possibly reduced single shot visibility , which might also be caused by multi mode effects in the initial double well @xcite .
exemplary interference measurements for a bec released from a double well trap are shown in figure [ fig_numberphase]b later in this manuscript .
the detection of number squeezing requires an accurate calibration of the deduced total atom number and the linearity of the imaging system .
simultaneously a high spatial resolution is often required @xcite .
to realize sufficiently large nonlinearities @xmath139 in the mesoscopic atom number regime a high atomic density and therefore tight confinement of the atoms is necessary , leading to optically think samples .
this complicates the use of in principle single atom sensitive fluorescence detection methods @xcite which suffer from light induced collision losses if the individual atoms are not spatially separated during the detection .
an alternative is to use absorption imaging @xcite in the high saturation intensity regime @xcite which provides good signal to noise ratio and high spatial resolution , however , its calibration is non trivial as outlined below . in high intensity absorption imaging
@xcite the full beer - lambert absorption formula is necessary to calculate the atomic column density @xmath182 from the counts per pixel @xmath183 on the picture containing the absorption information and on the reference picture @xmath184 ( the ccd camera pixels but it is mostly omitted below to ensure better readability ) . for the image analysis
the mean light intensity on the reference picture has to be normalized to the mean light intensity on the picture with the atomic signal . for its determination the area containing the atomic signal
is not taken into account .
not only the optical density @xmath185 but also the difference in the counts @xmath186 contribute to the signal : @xmath187 here @xmath188 is the linear extension of the ccd pixel taking the magnification into account , the resonant cross section , @xmath189 the clebsch - gordan coefficient , @xmath190 a correction factor obtained from a comparison to simulations as explained below and @xmath191 is the imaging pulse duration .
the factor @xmath192 contributes to the linear part of the formula where @xmath193 is the total quantum efficiency and @xmath194 the gain factor of the camera .
@xmath195 is the light angular frequency and @xmath196 the saturation energy of the transition .
the gain factor @xmath194 is determined by measuring the noise curve @xmath197 of the ccd camera .
it can be measured over the full dynamic range using an incoherent light source ( e.g. a led ) such that the light hitting the active region of the ccd is uncorrelated featuring shot noise limited noise characteristics .
the factor @xmath194 follows from the slope of the ccd s noise curve via a linear fit in the working region where the camera noise is not read out noise dominated .
the calibration of the @xmath198 factor is based on the quantum efficiency measurement which itself relies on the power meter calibration .
furthermore , for a typical numerical aperture around @xmath199 approximately @xmath200 of the solid angle is covered by the imaging lens such that a non negligible fraction of the scattered photons reaches the camera .
this effect is not taken into account in equation ( [ eq.imaging ] ) . a possible method to cross
check the inferred number of atoms is to prepare the bec in a very well known trap configuration and image it with very high imaging intensity ( typically @xmath201 ) . in this regime
only the linear part of the beer - lambert formula contributes , but the signal to noise ratio is poor .
imaging is done with a very short pulse ( @xmath202s ) to minimize diffusive broadening of the profile due to photon scattering .
the extracted average profile along an axis much bigger than the optical resolution is calculated for different correction factors @xmath190 and compared to theoretical profiles obtained from three dimensional gross - pitaevskii equation simulations varying the total numbers of atoms ( figure [ fig : calibration]a ) .
the correction factor is finally found by minimizing the quadratic deviation between the simulated and measured profile .
the expected corrections are small such that @xmath190 is should be close to unity .
[ t ] atoms yields a factor of two smaller slope while the absolute number is underestimated by only @xmath203 .
this results in a strong bias of the measured atom number fluctuations in this regime .
the reference atom number in this figure was measured for @xmath204s expansion time and with high imaging intensity @xmath205 ( black circles ) @xcite .
this figure was originally published in the supplementary online material to @xcite . ]
calibration of the imaging is completed by the measurement of the clebsch - gordan coefficient @xmath189 .
this is done by imaging the atoms with varying imaging intensities @xmath206 .
a correct clebsch - gordan factor means intensity independent atom number measurements ( see figure [ fig : calibration]b ) .
+ the total atom number follows from direct summation over the region where the atoms are detected ( not by a fit which might bias fluctuation measurements ) .
the size of this region is typically chosen to three standard deviations as obtained from a gaussian fit to the atomic cloud .
[ t ] [ t ] quantitative measurements of the atom number using absorption imaging is limited to atomic clouds with a size larger than the optical resolution of the imaging system .
for to small clouds the non - linear part of equation ( [ eq.imaging ] ) causes a systematic error in the detected atom number .
this can be easily understood in the limit where the optical resolution is much larger than the pixel size .
the detected column density @xmath207 per pixel is calculated from photon counts integrated over the size of the pixel @xmath208 where we omitted the linear term in equation ( [ eq.imaging ] ) .
correct atom number calculation @xmath209 however requires to calculate @xmath210 in general @xmath211 holds such that the calculated atom number is wrong . when the pixel size is smaller than the point spread function @xmath212 of the imaging system the same argument holds , but the main averaging effect is due to the convolution with the point spread function , e.g. the replacement @xmath213 has to be made , where @xmath214 means the convolution of the functions @xmath212 and @xmath215 .
figure [ fig : nonlinear_effects ] shows an example of this effect .
the nonlinearity in the atom number calculation for high optical densities and small atomic clouds is clearly visible .
it leads not only to an underestimation of the absolute atom number but also to an underestimation of the fluctuations in the high density region due to the smaller local slope .
by slightly expanding the atomic clouds before imaging this effect can be minimized .
the atoms are detected via their resonant interaction with the probe light whose noise characteristics add to the atomic noise of interest .
at least the shot noise of the detection light adds to the detected noise , but especially in absorption imaging interference fringes easily increase this minimal background noise level .
these fringes originate from motion of the imaging systems optics or air motion in the light path .
special care has to be taken to eliminate these effects .
the detection light shot noise contribution can be inferred pixel by pixel on both , the reference and absorption picture , using the camera noise calibration curve .
standard error propagation of equation ( [ eq.imaging ] ) allows for a conversion of the photonic variance @xmath216 into atomic variance : @xmath217 here the photon shot noise contribution @xmath218 is calculated by a expansion up to second order , assuming a gaussian distribution for the photon statistics @xmath219 .
photon shot noise on different pixels ( i , j ) is uncorrelated such that the total contribution in a given area of the ccd chip @xmath220 can be obtained by summation over the variance per pixel .
the photon shot noise is independent of the atomic noise @xmath221 .
thus , the best estimate of the true atomic fluctuations can be obtained by subtracting it from the detected fluctuations @xmath222 .
the uncertainty in the estimated amount of photon shot noise is typically in the order of a few percent .
+ in the strong saturation regime the imaging light intensity @xmath223 controls the transparency of the atomic cloud @xcite .
the imaging signal to noise ratio can be optimized by choosing the proper light intensity depending on the atomic density and the camera parameters . from the discussion above it is clear that correct calibration of the imaging system
is non - trivial .
therefore it is important to have an independent test of the deduced atomic variances .
one possibility is to hold an atomic sample in the trap until a large fraction of the atoms is lost by background gas collisions .
this uncorrelated one - body loss tends to restore a poissonian atom number distribution .
however , the poissonian limit is only reached asymptotically and the fluctuations might be super - poissonian for an initial distribution governed by experimental instabilities or slightly sub - poissonian if the loss is dominated by two or three body collisions @xcite .
+ when working with internal degrees of freedom coherent spin states can be prepared by precise electromagnetic coupling pulses .
their well known noise characteristics , that is , relative atom number fluctuations at the shot noise limit , can be directly used for the imaging calibration .
figure [ fig_imaginglinearity ] shows such a calibration check which reveals a correctly calibrated imaging over a large range of total atom numbers @xcite . in the discussion of the squeezing generation we assumed vanishing external noise during the evolution .
in real experiments , however , fluctuations of the differential energy shift @xmath111 coupling via the @xmath23 term in the hamiltonian ( [ eq.spin ] ) cause phase noise @xcite . both , the adiabatic and non - adiabatic approach to generate coherent spin squeezed states suffer from this noise problem .
especially the non - adiabatic method is very sensitive to dc phase noise ( fluctuations from shot to shot ) .
a spin echo technique @xcite can be used to effectively remove the sensitivity to these shot to shot fluctuations , however , the sensitivity to low but finite frequency fluctuations remains problematic .
in general the spectral sensitivity is dependent on the duration of the free evolution and knowing the noise spectrum a specially designed spin echo technique can be used to optimize the spectral sensitivity .
figure [ fig_phasenoise ] shows an exemplary sensitivity curve versus frequency of the environmental noise with and without spin echo .
[ t ] between optical dipole trap and optical lattice . in equilibrium
the chemical potential @xmath224 is the same in both wells which results in an occupation number difference between the left and right mode dependent on the differential energy shift @xmath225 between them . ] in the adiabatic case the final relative atom number fluctuations are also influenced by the noise in the differential energy shift , however , the main contribution is due to shot to shot fluctuations .
the continuous coupling of the two states results effectively in continuous spin echo pulses counteracting dephasing .
the remaining dc sensitivity can be understood intuitively for the example of the external double well system , but the argumentation also holds for an analogous internal system .
the system response to an imbalance in the offset energy for the two wells ( a dc differential shift ) is to adapt the relative atomic population such that the overall chemical potential is constant .
this means an increased population of the lower state such that the extra interaction energy balances the external energy offset .
an estimation for the noise sensitivity can be obtained by the following consideration where the main experimental challenge , the relative movement of the dipole traps that generate the trapping potential @xcite becomes obvious .
figure [ fig_trapshift ] illustrates this situation .
the optical dipole trap generates harmonic confinement in axial direction @xmath226 , where @xmath227 is the atomic mass .
fluctuations of the energy difference @xmath225 between the two potential minima separated by @xmath188 due to fluctuations of the barrier position @xmath228 is given by @xmath229 .
the overall chemical potential in local density approximation @xcite is @xmath230 and in equilibrium it is the same for the two wells .
the contribution due to inter atomic interactions @xmath231 balances the change in @xmath232 : @xmath233 , where we used @xmath234 @xcite and @xmath235 . to give an estimation of the required position stability of the trapping potentials we calculate the actual numbers for the trapping parameters of the double well trap in reference @xcite .
the nonlinearity is @xmath236hz and the longitudinal trapping frequency @xmath237hz .
we obtain @xmath238 for the position fluctuations leading to an extra noise of the same order as the shot noise limit @xmath239db for @xmath240 atoms .
since these technical induced fluctuations add to the variance caused by the atomic quantum state , their magnitude has to be much smaller than the shot noise level in order to measure a reasonable amount of number squeezing . as a figure of merit ,
position fluctuations of @xmath241 nm between different experimental realizations limit the best observable number squeezing to @xmath242db .
this corresponds to roughly one percent of the double well spacing in this experiment @xcite .
[ t ] the influence of finite temperature is very different in the two settings discussed here .
while the effective temperature in the spin degree of freedom is nearly zero in the internal case , finite temperature effects easily dominate the fluctuations when dealing with a bec in a double well trap @xcite .
the difference emerges from the very different initial energy gap between the two modes comprising the pseudo spin . when using two hyperfine states their energy separation is typically in the order of several gigahertz corresponding to temperatures of a few hundred millikelvin , much larger than the temperature in the bec of a few tens of nanokelvin .
thus , using almost noise free coupling fields and short coupling pulses much faster than any thermalization timescale , spin states with almost zero temperature can be realized ( see references @xcite for the effect of finite temperature in this system . ) .
[ t ] the situation is very different in the external double well potential , where the pseudo spin in comprised of the two center of mass modes in the two potential wells . in a unbiased ( @xmath148 ) situation
, there is no or a negligibly small energy gap separating the two modes such that several collective modes are excited even at nanokelvin temperatures .
this leads to non pure initial spin states and to increased spin fluctuations as compared to a coherent spin state .
nevertheless the adiabatic spin squeezing technique results in an effective cooling of the spin fluctuations such that coherent spin squeezed states can be realized @xcite .
figure [ fig_adiabaticisothermal ] shows the number fluctuations and mean spin lengths for different initial temperatures comparing the adiabatic method to the isothermal case .
parameters are the same as for the experiments reported in @xcite .
note that the fluctuations for the ground state would be orders of magnitude better for the same hamiltonian parameters , showing the dominating effects of finite temperature .
in the previous sections we took two settings as examples for the generation of adiabatic and non - adiabatic spin squeezing in a bec . we will briefly discuss some results of these experiments @xcite in the following .
a more detailed description of these experiments can be found in @xcite .
the experiments discussed in this section have been originally published in @xcite and in @xcite .
approximately @xmath240 condensed rubidium atoms were loaded into a symmetric double trap where initially the potential barrier separating the two modes was small . for the initial parameters the system was already in the josephson regime due to the rather strong nonlinearity .
next the potential barrier was increased adiabatically , driving the system deeper into the josephson regime .
the decrease of the tunnel coupling resulted in a decrease of the relative number fluctuations directly corresponding to the spin fluctuations in @xmath34 direction .
the fluctuations were redistributed to the orthogonal spin direction such that the mean spin length decreased . the simultaneous change in the number fluctuations and
mean spin length can be seen directly on the raw data ( figure [ fig_numberphase ] ) .
predicting the optimal time required for an adiabatic change of the parameters is not trivial in such an experiment , especially because additional aspects such as particle losses or heating limit the duration of the experiments .
therefore , the number squeezing was measured versus ramp up time of the double well potential while the final barrier height was kept constant .
figure [ fig_adiabaticity ] shows the results of this experiment . in order to explain the dependence of number squeezing versus ramp up time both finite temperature and finite fluctuations of the trap position had to be taken into account .
they dominate the observed fluctuations ( details of this experiment can be found in @xcite ) . around the feshbach resonance @xcite .
the black solid line is a fit to the real part of @xmath245 @xcite revealing an inelastic width of @xmath246 mg in accordance with the theoretical prediction @xcite .
the elastic width extracted from the fit is @xmath247 mg .
the center of the resonance is at @xmath248 g , but the absolute magnetic field has an uncertainty of approximately @xmath249 mg in this measurement @xcite . the background scattering length @xmath250 was chosen as a free parameter and the fit reveals @xmath251 . ] despite of these limitations coherent spin squeezed states in the double well trap were realized with a maximal squeezing factor of @xmath252db @xcite .
however , the implementation of a full atom interferometer beyond the standard quantum limit remains an experimental challenge with such a setup even though the feasibility of the production of coherent spin squeezed states was demonstrated .
the analog of electromagnetic coupling pulses rotating the spin state non - adiabatically needs to be implemented in order to realize a ramsey sequence with the spin squeezed state as input state .
the required non - adiabatic changes are difficult to realize without the possibility to change the nonlinearity directly ( which could be done by using a feshbach resonance @xcite ) since fast changes in the external trapping potential lead to unwanted excitation of the condensate .
+ using internal hyperfine states to define the two modes instead of the center of mass modes these limitations can be overcome as presented in the following section .
when working with internal hyperfine states of the atoms fast and precise spin rotations are readily available using electromagnetic coupling .
the experimental challenge is to obtain a sufficiently high nonlinearity such that spin squeezed states can be realized within the experimental coherence time .
several ideas to increase the nonlinear interaction have been experimentally realized reaching from cavity mediated interactions in thermal gases @xcite to state dependent microwave traps on atom chips which allow for the control of the spatial overlap of the two spin modes @xcite .
we focus here on another method exploiting an interstate feshbach resonance between two hyperfine states of rubidium around @xmath253 g @xcite and @xmath254 hyperfine states . ] .
such a feshbach resonance may be used to increase the nonlinearity but at the cost of increased particle losses @xcite .
figure [ fig_feshbach ] shows the measured resonantly enhanced effective nonlinearity @xcite .
the simultaneously enhanced losses limit the available time to generate coherently spin squeezed states to a few tens of milliseconds favoring the fast non - adiabatic `` one axis twisting '' method proposed in @xcite for the generation of spin squeezing @xcite .
coherent spin squeezed states with inferred squeezing factors up to @xmath255db could be realized within an evolution time of @xmath256ms @xcite .
the use of a spin echo technique and an actively stabilized magnetic field @xcite was crucial here in order to minimize the dephasing effects .
given the possibility of precise spin rotations a full interferometric sequence can be implemented . in a prototypical measurement
enhanced interferometric sensitivity of @xmath203 beyond the standard quantum limit has been demonstrated @xcite using coherent spin squeezed states within the ramsey interferometer .
figure [ fig_interferometer ] shows a direct comparison of the ramsey interferometer performance for coherent spin states as input state versus coherent spin squeezed states . in this direct comparison
the performance increase due to the spin squeezing is even larger ( @xmath257 ) .
however , it should be made clear that experimental noise is included here , that is , the comparison is not relative to the standard quantum limit .
measured is the increase in performance due to spin squeezing in this particular interferometric setup .
several other experiments recently succeeded in the realization of quantum enhanced interferometers .
an interferometer with @xmath258db precision gain has been realized using a cold sample of cesium atoms where coherent spin squeezing has been generated by quantum non - demolition measurements @xcite . using cavity mediated interactions in a cold rubidium gas to generate coherent spin squeezing the time dependent performance of a similar interferometer has been characterized in terms of the allan variance @xcite .
these different experiments show that atomic coherent spin squeezing can nowadays be realized with a variety of techniques and in quite different systems .
it remains an open challenge to realize a large amount of coherent spin squeezing with atom numbers comparable to those used in modern linear interferometers .
however , the basic concept of spin squeezing and its experimental feasibility has been demonstrated such spin squeezing might help to push the limit of precision measurements in the future .
i thank the bec team at the kirchhoff institute for physics in heidelberg for their support and practical help .
especially i thank markus oberthaler for his great support and for his supervision of my phd thesis .
special thanks goes to andreas weller , eike nicklas , tilman zibold and jrme estve for the joint work on our squeezing experiments . | squeezed states , a special kind of entangled states , are known as a useful resource for quantum metrology . in interferometric sensors they allow to overcome the `` classical '' projection noise limit stemming from the independent nature of the individual photons or atoms within the interferometer . motivated by the potential impact on metrology as wells as by fundamental questions in the context of entanglement , a lot of theoretical and experimental effort has been made to study squeezed states .
the first squeezed states useful for quantum enhanced metrology have been proposed and generated in quantum optics , where the squeezed variables are the coherences of the light field . in this tutorial
we focus on spin squeezing in atomic systems .
we give an introduction to its concepts and discuss its generation in bose - einstein condensates .
we discuss in detail the experimental requirements necessary for the generation and direct detection of coherent spin squeezing .
two exemplary experiments demonstrating adiabatically prepared spin squeezing based on motional degrees of freedom and diabatically realized spin squeezing based on internal hyperfine degrees of freedom are discussed . |
the masses of active neutrinos are the most prominent evidence for physics beyond the standard model .
the simplest and sufficient extension is to introduce right - handed neutrinos ( the seesaw mechanism @xcite ) .
they can also explain the baryon asymmetry of the universe @xcite .
right - handed neutrinos have been searched for a long time at various experiments and observations ( for compilations , see refs .
their detectability is determined by two parameters : masses and mixings . in this letter
, we derive upper and lower bounds on the mixing of the lightest right - handed neutrino .
the lower bound comes from the seesaw relation , and the upper one is from the perturbativity . to obtain the upper bound , we take into account the cancellation between the contributions of right - handed neutrinos to the mass matrix of active neutrinos .
we also consider the renormalization group evolution ( rge ) of yukawa couplings .
the new upper bound is stronger than the existing bounds by orders of magnitude for many cases .
this bound is crucial for future collider searches @xcite .
the perturbativity of the neutrino yukawa couplings without rge was discussed in ref .
the rge effect was partially studied in the context of grand unification @xcite .
note that stronger bounds may be derived by the vacuum stability condition @xcite , but it largely depends on the top yukawa coupling and the details of the scalar sector .
here we focus on the perturbativity to be conservative .
our results do not change much even if we assume the stability .
we consider the standard model extended by right - handed neutrinos @xmath2 .
the neutrino masses are induced by the lagrangian @xmath3 @xmath4 are yukawa couplings of neutrinos and higgs doublet @xmath5 .
we assume @xmath6 to realize the seesaw mechanism .
the mass eigenstates of neutrinos are three active neutrinos @xmath7 ( @xmath8 ) with masses @xmath9 and @xmath10 heavy neutral leptons ( hnls ) @xmath11 with masses @xmath12 .
the labeling of hnls is defined as @xmath13 .
the masses and mixings are related by the seesaw relation @xmath14 where @xmath15 denotes the active neutrino mixing matrix , @xmath16 is the mixing element of hnl @xmath11 , @xmath17 and @xmath18 . the relation ( [ eseesaw ] )
is modified by radiative corrections @xcite .
these corrections do not change our final results since they give subleading contributions .
the yukawa couplings which satisfy eq .
( [ eseesaw ] ) are convenient to be parametrized as @xcite @xmath19 where @xmath20 is a @xmath21 complex matrix which is not determined by neutrino oscillation experiments .
it satisfies @xmath22 for @xmath23 and @xmath24 diag(0,1,1 ) or diag(1,1,0 ) for @xmath25 with the normal hierarchy ( nh ) or inverted hierarchy ( ih ) .
the mixing of the lightest hnl is expressed as @xmath26 cases , @xmath27 where @xmath28 and @xmath29 are real and imaginary parts of @xmath20 .
@xmath30 is @xmath31 for @xmath32 with the nh ( ih ) and the lightest neutrino mass for @xmath33 . in the first equality ,
we have used @xmath34 , which holds for @xmath35 .
the last equality holds when @xmath36 . for @xmath37
, there is no lower bound since we can choose @xmath38=0 , which means that @xmath39 decouples from the seesaw mechanism .
if the lightest neutrino is massless , @xmath39 can also decouple for @xmath33 .
from eq . ( [ eth1 ] ) , we can see that the mixing @xmath40 can be much larger than @xmath41 if @xmath42 is much larger than 1 .
actually , such @xmath20 is needed to discover the hnl by collider experiments .
for example , the bound from @xmath43 boson decay is @xmath44 for @xmath45 gev @xcite , and then @xmath46 for @xmath47 . in this case , the contribution of @xmath39 in the right - hand side of eq .
( [ eseesaw ] ) need to be mostly canceled by those of heavier hnls ( @xmath20 can not be diagonal ) .
the mixing , however , can not be arbitrarily large , since yukawa couplings are constrained by perturbativity as discussed below .
the diagonal components of the seesaw relation ( [ eseesaw ] ) are @xmath48 in this section , we assume that @xmath39 is detectable at colliders , or @xmath49 for at least one of @xmath50 .
the seesaw relation ( [ ess ] ) and the triangle inequality give @xmath51 .
the perturbativity bound can be estimated as @xmath52 as the perturbativity conditions .
note that the bound gets more severe when the mass hierarchy is stronger @xmath53 .
in such a case , the perturbativity conditions on @xmath54 ( @xmath55 ) give the stronger bound on @xmath56 than the perturbativity of @xmath57 .
we can use eq .
( [ etha1 ] ) to obtain a lower bound on the mass of the second lightest hnl when @xmath58 and @xmath59 are measured by future experiments .
if @xmath60 hnls decouples from the seesaw mechanism ( like a dark - matter - candidate hnl ) , eq .
( [ etha1 ] ) can be slightly modified by @xmath61 .
we have to take into account the rge of yukawa couplings to keep perturbativity at high energy scale .
we consider two right - handed neutrinos ( @xmath25 ) for the calculation of rge below . for @xmath23 , some structure of @xmath20
has to be assumed , but the bound would change only by an @xmath62 factor [ see eq .
( [ etha1 ] ) ] .
it is useful to calculate the rge of neutrino yukawa couplings in the basis that the active neutrino mass matrix is diagonal , or @xmath63 . in this basis ,
our result do not depend on the unknown dirac and majorana phases in @xmath15 because contributions of charged leptons to the rge are negligible .
the mixings @xmath64 do not depend on the choice of the basis .
a typical example of yukawa coupling rge is shown in fig .
we have used one - loop renormalization group equations @xcite .
the input parameters are taken to the central values in refs .
@xcite . in our basis of eq .
( [ eq : casas ] ) , @xmath65 for the nh .
the absolute values of the other neutrino yukawa couplings are monotonically increasing , so we define the perturbativity condition here as @xmath66 is a renormalization scale and @xmath67 gev ( the difference of the results from those applying @xmath68 is negligible ) .
the perturbativity of the top yukawa coupling is satisfied since it is smaller than the maximal component of @xmath69 at high energy scales .
renormalization group evolution of top and neutrino yukawa couplings @xmath70 and @xmath71 in the case with @xmath72 , @xmath73 and the nh.,width=491 ] constraints on the mixing @xmath40 is shown in fig .
[ fseesaw ] .
we have drawn the perturbativity bound with @xmath74 for both the nh and ih cases , but the difference is negligible .
the perturbativity bound falls as @xmath75 when @xmath74 and @xmath76 when @xmath77 is fixed [ see eq .
( [ etha1 ] ) ] .
we can see that two hnls can not be heavier than @xmath78 gev .
it should be noted that the seesaw lower bound eq .
( [ eq : low ] ) can be weaker or stronger for @xmath79 depending on the lightest neutrino mass .
constraints on the mixing of the lightest hnl .
shaded regions are excluded .
the upper bounds from the perturbativity are shown by red solid lines .
the bound eq .
( [ etha1 ] ) with @xmath80 is shown by the red dot - dashed line labeled `` no rge '' .
blue dashed lines are the bounds eq .
( [ eq : low ] ) for the nh and ih .
dotted lines show experimental and cosmological bounds and future sensitivities @xcite .
, width=415 ] the low mass region is depicted in fig .
[ fzoom ] with the constraints from various experiments ( see refs . @xcite and references therein ) .
the constraints are mostly dominated by those on @xmath81 .
the perturbativity bound can be much stronger than those of past experiments , and also comparable to the sensitivities of future experiments .
low mass region of fig .
[ fseesaw ] .
the labels of dotted lines indicate names of experiments or methods @xcite . ,
width=415 ] note that fig .
[ fzoom ] says that low energy experiments can constrain hnls much heavier than the electroweak scale , as indicated by eq .
( [ etha1 ] ) .
the constraint is not much changed by @xmath82 [ see eq . ( [ etha1 ] ) ] .
if @xmath83 gev , the mixing @xmath40 have to be smaller than the line labeled `` @xmath84 gev '' .
for example , if a hnl with @xmath85 gev and @xmath86 is found , the second lightest hnl must be lighter than @xmath87 gev by the perturbativity and the seesaw relation .
it means that thermal leptogenesis @xcite , which requires two hnls heavier than @xmath88 gev to explain the baryon asymmetry of the universe @xcite , is disfavored for @xmath33 . in this case
, the baryon asymmetry have to be generated by another mechanism , such as resonant leptogenesis @xcite or baryogenesis via neutrino oscillations @xcite .
we have explicitly shown two constraints on right - handed neutrinos .
the lower bound comes from the seesaw relation , and the upper one comes from perturbativity .
the mixing of the lightest hnl is related to those of heavier hnls via the seesaw mechanism .
if the lightest hnl is found by experiments , the second lightest hnl have to be sufficiently light .
this fact can be used to probe high energy phenomena such as leptogenesis .
t.a . was supported by jsps kakenhi grants no . 25400249 , no .
26105508 , and no . 15h01031 .
p. minkowski , phys .
b * 67 * , 421 ( 1977 ) ; t. yanagida , in _ proceedings of the workshop on unified theory and baryon number of the universe _ , edited by o. sawada and a. sugamoto ( kek , tsukuba , 1979 ) ; m. gell - mann , p. ramond , and r. slansky , in _ supergravity _ , edited by p. van niewenhuizen and d. freedman ( north holland , amsterdam , 1979 ) [ arxiv:1306.4669 [ hep - th ] ] ; s. l. glashow , nato sci .
b * 61 * , 687 ( 1980 ) ; r. n. mohapatra and g. senjanovic , phys .
lett . * 44 * , 912 ( 1980 ) .
m. fukugita and t. yanagida , phys . lett .
b * 174 * ( 1986 ) 45 .
a. y. smirnov and r. zukanovich funchal , phys .
d * 74 * , 013001 ( 2006 ) [ hep - ph/0603009 ] .
a. atre , t. han , s. pascoli and b. zhang , jhep * 0905 * , 030 ( 2009 ) [ arxiv:0901.3589 [ hep - ph ] ] .
f. f. deppisch , p. s. bhupal dev and a. pilaftsis , new j. phys .
* 17 * , no . 7 , 075019 ( 2015 ) [ arxiv:1502.06541 [ hep - ph ] ] . c. adams _ et al . _
[ lbne collaboration ] , arxiv:1307.7335 [ hep - ex ] .
a. blondel _ et al .
_ [ fcc - ee study team collaboration ] , arxiv:1411.5230 [ hep - ex ] .
a. kobach and s. dobbs , phys .
d * 91 * , no .
5 , 053006 ( 2015 ) [ arxiv:1412.4785 [ hep - ph ] ] .
s. antusch and o. fischer , jhep * 1505 * , 053 ( 2015 ) [ arxiv:1502.05915 [ hep - ph ] ] .
s. banerjee , p. s. b. dev , a. ibarra , t. mandal and m. mitra , phys .
d * 92 * , 075002 ( 2015 ) [ arxiv:1503.05491 [ hep - ph ] ] .
j. a. casas , j. m. moreno , n. rius , r. ruiz de austri and b. zaldivar , jhep * 1103 * , 034 ( 2011 ) [ arxiv:1010.5751 [ hep - ph ] ] .
t. tsuyuki , ptep * 2015 * , 011b01 ( 2015 ) [ arxiv:1411.2769 [ hep - ph ] ] .
j. elias - miro , j. r. espinosa , g. f. giudice , g. isidori , a. riotto and a. strumia , phys .
b * 709 * , 222 ( 2012 ) [ arxiv:1112.3022 [ hep - ph ] ] .
w. rodejohann and h. zhang , jhep * 1206 * , 022 ( 2012 ) [ arxiv:1203.3825 [ hep - ph ] ] .
w. grimus and l. lavoura , phys . lett .
b * 546 * , 86 ( 2002 ) [ hep - ph/0207229 ] .
d. aristizabal sierra and c. e. yaguna , jhep * 1108 * , 013 ( 2011 ) [ arxiv:1106.3587 [ hep - ph ] ] .
p. abreu _ et al .
_ [ delphi collaboration ] , z. phys .
c * 74 * , 57 ( 1997 ) [ z. phys .
c * 75 * , 580 ( 1997 ) ] .
b. grzadkowski and m. lindner , phys .
b * 193 * , 71 ( 1987 ) .
k. a. olive _ et al .
_ [ particle data group collaboration ] , chin . phys .
c * 38 * , 090001 ( 2014 ) .
k. hamaguchi , h. murayama and t. yanagida , phys .
d * 65 * , 043512 ( 2002 ) [ hep - ph/0109030 ] .
s. davidson and a. ibarra , phys .
b * 535 * , 25 ( 2002 ) [ hep - ph/0202239 ] .
see , for example , reviews ; g. f. giudice , a. notari , m. raidal , a. riotto and a. strumia , nucl .
b * 685 * , 89 ( 2004 ) [ hep - ph/0310123 ] ; w. buchmuller , r. d. peccei and t. yanagida , ann .
rev . nucl . part .
* 55 * , 311 ( 2005 ) [ hep - ph/0502169 ] .
a. pilaftsis , phys .
d * 56 * , 5431 ( 1997 ) [ hep - ph/9707235 ] .
e. k. akhmedov , v. a. rubakov and a. y. smirnov , phys . rev .
lett . *
81 * , 1359 ( 1998 ) [ hep - ph/9803255 ] .
t. asaka and m. shaposhnikov , phys .
b * 620 * , 17 ( 2005 ) [ hep - ph/0505013 ] . | we consider the standard model extended by right - handed neutrinos to explain massive neutrinos through the seesaw mechanism .
the new fermion can be observed when it has a sufficiently small mass and large mixings to left - handed neutrinos .
if such a particle is the lightest right - handed neutrino , its contribution to the mass matrix of active neutrinos needs to be canceled by that of a heavier one .
yukawa couplings of the heavier one are then larger than those of the lightest one .
we show that the perturbativity condition gives a severe upper bound on the mixing of the lightest right - handed neutrino , depending on the masses of heavier ones .
models of high energy phenomena , such as leptogenesis , can be constrained by low energy experiments .
perturbativity in the seesaw mechanism * * 0.5 cm takehiko asaka@xmath0 , takanao tsuyuki@xmath1 + 0.4 cm _ @xmath0department of physics , niigata university , 950 - 2181 niigata , japan _
+ _ @xmath1graduate school of science and technology , niigata university , niigata 950 - 2181 , japan
_ 0.2 cm ( septemter 9 , 2015 ) .5 cm |
it has been proposed that the brain is in or near a critical state | its dynamics may be positioned at the border between spatiotemporal order and disorder , reminiscent of non - equilibrium phase transitions in thermodynamic systems .
the concept of brain criticality is attractive because critical systems display optimal performance on several characteristics such as information transfer , wide dynamic range , information capacity @xcite , and long - term stability .
criticality relates closely to self - organization , which is considered crucial to cortical functioning .
a hallmark of critical behavior is the presence of power laws .
power laws symbolize scale - free behavior , adopting the same form on all time scales : they are self - similar .
consider the case of a scale - free auto - correlation function @xmath0 .
the corresponding power law obeys the form @xmath1 , i.e. if time @xmath2 is rescaled to @xmath3 , then the shape of @xmath0 is preserved and only rescaled by a factor @xmath4 .
the scaling exponent @xmath5 is referred to as the hurst exponent @xcite and qualifies the underlying correlation structure : @xmath6 corresponds to an uncorrelated , random process whereas @xmath7 indicates persistent , long - range correlations .
there is accumulating evidence for the presence of power laws in brain activity @xcite .
neural spikes come in avalanches that display scale - free distributions .
spectral distributions of encephalographic signals have @xmath8-structures and their auto - correlation structures also show power - law behavior @xcite . in the spatial domain , scale - free distributions
have been observed in functional as well as neuroanatomical connectivity patterns .
previous work has focused only on spatially local measures of brain activity or considered pairs of nodes in networks @xcite rather than analyzing global brain activity . in complex systems , the global activity can be very informative about the generating dynamical structure @xcite , in particular when studying critical behavior @xcite . here
, we adopted these concepts to investigate criticality in the brain . by using source - reconstructed magneto - encephalographic ( meg ) signals we sought to disambiguate between the scaling characteristics of amplitude and phase fluctuations because they may resemble different aspects of cortical functioning .
magneto - encephalographic ( meg ) signals of ten subjects were recorded and sampled at 1 khz in eyes - closed resting state for approximately five minutes . after down sampling to 250 hz
, signals were beamformed onto a ninety - node brain parcellation , yielding ninety time series @xmath9 per subject .
data were previously published by cabral and coworkers @xcite .
signals @xmath9 were filtered with a second - order iir - bandpass filter in the alpha band ( 8 - 12 hz ) and ( upper ) beta band ( 20 - 30 hz ) . with the hilbert
transform we constructed the analytic signal and defined phase @xmath10 and amplitude @xmath11 as functions of time @xmath12 ; @xmath13 indexes either the alpha or the beta frequency band .
we used two collective variables to capture whole - brain activity per subject .
first , we defined the phase synchronization , @xmath14 and , second , the mean amplitude , @xmath15 we note that @xmath16 is the modulo of the complex - valued kuramoto order parameter @xcite .
z_-scored @xmath17 and @xmath18 to reduce between - subject variability such that we could assess subject - averaged behavior by means of a detrended fluctuation analysis ( dfa , @xcite ) , as described below . to relate our study to the previously established results in rsns
, we also examined the amplitude dynamics in more detail ( see also @xcite ) .
the expression of rsns is mainly reflected in the low - frequency content of the amplitudes @xmath11 , whose time scale is comparable to those of the blood - oxygenation - level - dependent ( bold ) signal @xcite . in order to study these slow amplitude dynamics
, we evaluated the longer time scales of the @xmath11 dynamics by extracting its amplitude @xmath19 and phase @xmath20 .
subsequently , we defined the collective variables associated with the amplitude dynamics : @xmath21 further analysis was identical to that described above ; see * figure [ fig4:fig1 ] * for illustration . + after filtering in the alpha band and corresponding hilbert - amplitude @xmath11 ( upper left ) and hilbert - phase @xmath22 ( lower left ) . for clarity
we decreased the time axis range and rescaled the @xmath10 trace in the lower left panel .
the amplitude is also displayed ( black ) in the upper middle panel together with its envelope @xmath23 ( red ; see main text for the detailed definition ) .
the hilbert - phase @xmath10 , equal to @xmath10 in the lower left panel , and the hilbert - phase of @xmath11 , @xmath20 , are displayed in the lower middle panel .
different slopes indicate separate time scales ( frequencies ) .
the upper right panel shows amplitudes @xmath18 and @xmath24 , the lower right one the phase order parameters @xmath25 and @xmath26,scaledwidth=95.0% ] dfa is considered robust against non - stationarity rendering it suitable for analyzing the temporal autocorrelation structure of encephalographic activity , in general , and global amplitude and phase synchronization , in particular .
we employed a modified form of dfa including ( bayesian ) model selection to verify the presence of power - law behavior . in a nutshell , to quantify the autocorrelation structure of ( the cumulative sum of ) a signal @xmath27 , one divides it into non - overlapping segments @xmath28 , with @xmath29 being discrete time steps and @xmath30 indexing the segments ; @xmath31 is the number of non - overlapping segments of length @xmath32 . in each segment
the linear trend @xmath33 is removed providing an estimate of fluctuations in terms of @xmath34 this definition yields a set of ` realizations ' of fluctuations @xmath35 that , in the presence of a power law , scale like @xmath36 , which is equivalent to @xmath37 . that is , in a log - log representation these fluctuations have a linear relationship with segment size .
dfa seeks to identify the scaling exponent @xmath38 that provides an estimate for the aforementioned hurst exponent @xmath5 . instead of computing the mean value of @xmath35 as in conventional dfa @xcite , we here determined the probability density function @xmath39 for every segment length @xmath32 ( see * figure [ fig4:fig2 ] * ) .
this approach allows for quantifying the appropriateness of a model @xmath40 for fitting the fluctuation structure @xmath41 as a function of @xmath42 by means of the log - likelihood function @xmath43 , where the tilde indicates a transformation to logarithmic coordinates . here , the model @xmath40 , parametrized by the set @xmath44 , may obey any arbitrary form including the linear one , which corresponds to a power law .
we tested this linear relationship against a set of alternative models ( * table [ tab4:tab1 ] * ) using both the bayesian information criterion and the akaike information criterion ( bic and aic@xmath45 , respectively ) .
the model resulting in the least value of the information criterion was selected as the proper model . whenever this yielded the linear model @xmath46 , we considered power - law behavior to be present and identified the scaling exponent with its slope .
+ p0.4 l @xmath47 & @xmath48 + @xmath49 & @xmath50 + @xmath51 & @xmath52 + @xmath53 & @xmath54 + @xmath55 & + since we were interested in the subject - averaged scaling exponents , we determined @xmath56 for every subject individually and averaged over subjects to obtain @xmath57 .
these averaged probability density functions were used both for model selection and to determine the scaling exponent @xmath38 .
the scaling range was given by @xmath58 = \left [ 1.875 \cdot 10^{2},~1.875 \cdot 10^{4 } \right ] \simeq \left [ 0.75,\ 75 \right]$ ] seconds , i.e. it spanned two decades .
in this range we used one hundred equally spaced window sizes ( on a logarithmic scale ) .
+ bootstrapping served to establish statistical significance using three types of surrogate data .
two of them consisted of randomly permuting temporal order whereas the third one only influenced cross - correlation structure . of all types we constructed 1000 surrogates and significance values were obtained using @xmath59 , where @xmath60 denotes the surrogate distribution and @xmath38 the obtained empirical value of the scaling exponent .
for the first type of surrogates , we randomly permuted the order parameter time series for each subject . with this
we evaluated our dfa and fitting procedure , since the surrogate time series lacked any temporal correlation structure and therefore should result in @xmath61 . with the second type of surrogate we evaluated whether the filtering procedure and hilbert analysis could have biased the results .
for this , we permuted all original time series @xmath62 . by permuting @xmath62 ,
all temporal structure was destroyed and therefore this constituted a rather weak null . in the third type of surrogates we performed a random cycling of @xmath63 by shifting the time indices of @xmath62 for each @xmath64 , but keeping their order intact . in this way we retained the original auto - correlation structure of @xmath65 and @xmath66 but destroyed the cross - correlations .
subsequent analyses for all collective variables were identical to those for the original data .
the presence of power - law scaling was evidenced by clear linear relationships of @xmath18 and @xmath25 fluctuations in log - log scale ; see * figure [ fig4:fig3]*. this was confirmed by the aic@xmath45 and bic values preferring the linear model in all cases ( * tables [ tab4:tab2]-[tab4:tab3 ] * ) . in both chosen frequency bands , the brain s network dynamics , as measured by the collective variables @xmath18 and @xmath25 , thus appeared to exhibit scale - free correlations over a very broad range of time scales .
+ in the alpha band the scaling exponents were 0.88 and 0.62 , and in the beta band 0.80 and 0.57 , for amplitude @xmath67 and phase synchronization @xmath68 , respectively .
that is , both @xmath18 and @xmath25 showed persistent behavior but amplitude had increased perseverance compared to phase ( * figure [ fig4:fig3 ] * ) .
we also found long - range temporal correlations in the variables @xmath69 and @xmath70 as shown in * figure [ fig4:fig4 ] * and confirmed by the aic@xmath45 and bic values by preferring the linear model , except in case of the aic@xmath45 for @xmath71 in the alpha band ( * tables [ tab4:tab2]-[tab4:tab3 ] * ) .
the difference in scaling behavior between @xmath69 and @xmath70 was similar to that between @xmath18 and @xmath25 .
+ p0.2 r r r r @xmath72 & 0.00 ( 0.00 ) & 0.00 ( 0.00 ) & 0.00 ( 0.00 ) & 0.00 ( 0.10 ) + @xmath73 & 5.85 ( 5.85 ) & 3.56 ( 3.56 ) & 4.92 ( 4.92 ) & 7.17 ( 7.27 ) + @xmath74 & 4.60 ( 2.12 ) & 4.60 ( 2.12 ) & 4.60 ( 2.12 ) & 2.38 ( 0.00 ) + @xmath75 & 16.21 ( 16.21 ) & 12.09 ( 12.09 ) & 16.95 ( 16.95 ) & 17.60 ( 17.70 ) + @xmath76 & 4.60 ( 2.12 ) & 4.50 ( 2.02 ) & 4.58 ( 2.10 ) & 2.48 ( 0.10 ) + @xmath77 & 4.89 ( 2.41 ) & 4.94 ( 2.46 ) & 5.08 ( 2.60 ) & 2.88 ( 0.50 ) + @xmath78 & 9.02 ( 4.11 ) & 7.01 ( 2.10 ) & 8.29 ( 3.38 ) & 6.63 ( 1.82 ) + @xmath79 & 4.60 ( 2.12 ) & 4.68 ( 2.20 ) & 4.60 ( 2.12 ) & 4.67 ( 2.30 ) + @xmath80 & 32.05 ( 29.57 ) & 6.01 ( 3.53 ) & 13.98 ( 11.50 ) & 3.69 ( 1.32 ) + testing against surrogate data confirmed the significance of these correlations .
* figure [ fig4:fig5 ] * depicts the scaling exponent distributions corresponding to the third type of surrogates . for all variables in both frequency bands
the scaling exponents significantly exceeded those of the surrogates ( @xmath81 ) .
+ to further highlight the peculiar role of amplitude and phase , we finally contrasted our results with the scaling of fluctuations of the mean meg activity .
for this , we applied our dfa to @xmath82 , i.e. we considered not the amplitude and phase but the ` raw ' meg signals .
this mean activity did not display long - range correlations but rather anti - persistent ones ( @xmath83 , * figure [ fig4:fig6 ] * ) .
p0.2 r r r r @xmath72 & 0.00 ( 0.00 ) & 0.00 ( 0.00 ) & 0.00 ( 0.00 ) & 0.00 ( 0.10 ) + @xmath73 & 0.37 ( 0.37 ) & 2.72 ( 2.72 ) & 0.35 ( 0.35 ) & 1.36 ( 1.36 ) + @xmath74 & 4.04 ( 1.57 ) & 4.60 ( 2.12 ) & 4.17 ( 1.69 ) & 4.50 ( 2.02 ) + @xmath75 & 6.29 ( 6.29 ) & 11.46 ( 11.46 ) & 5.33 ( 5.33 ) & 6.83 ( 6.83 ) + @xmath76 & 3.99 ( 1.51 ) & 4.60 ( 2.12 ) & 4.10 ( 1.62 ) & 4.49 ( 2.01 ) + @xmath77 & 4.20 ( 1.72 ) & 4.57 ( 2.09 ) & 4.33 ( 1.85 ) & 4.63 ( 2.15 ) + @xmath78 & 8.15 ( 3.24 ) & 9.18 ( 4.27 ) & 8.36 ( 3.44 ) & 8.89 ( 3.98 ) + @xmath79 & 4.02 ( 1.54 ) & 4.62 ( 2.14 ) & 4.14 ( 1.66 ) & 4.52 ( 2.04 ) + @xmath80 & 55.24 ( 52.76 ) & 10.34 ( 7.86 ) & 26.24 ( 23.76 ) & 12.97 ( 10.49 ) + +
we report power - law scaling in both amplitude and phase of collective neural activity on long time scales , which is consistent with the hypothesis that the brain operates in a critical state .
operating in a critical state is not the only way a system can generate power - law scaling .
systems in subcritical states @xcite or merely stochastic systems may also display power laws .
biological systems display sub- and supercritical dynamics but they can be tuned into criticality .
we favor the interpretation of critical states , also because it is consistent with scale - free auto - correlation structures of single channel eeg activity @xcite , size and duration of neural avalanches and , in the spatial domain , degree distributions of neuroanatomical and functional connectivity networks .
such scaling laws in neuronal dynamics are also correlated with those found in behavior @xcite .
previous studies addressing power - law scaling in neuronal dynamics were solely based on spatially local measures .
for instance , power - law scaling has been reported using pair - wise synchronization measures like pli and @xmath84 by , e.g. , kitzbichler and coworkers @xcite and farmer @xcite .
in contrast , we evaluated scaling behavior on a global brain scale by using overall amplitude and phase synchrony as collective variables . analyzing
the scaling behavior of these variables enabled us to directly compare phase and amplitude behavior in brain activity . what does the difference in power - law scaling of amplitudes and phases tell us about information processing in the brain ?
despite the fact that amplitude and phase differ in their perseverance , both show long - range correlations over a scale of hundreds of seconds , suggesting that they reflect ` memory ' of cortical states .
a higher scaling exponent reflects a slower decay of auto - correlations and hence a more predictable signal with decreased entropy .
therefore , we speculate that the difference in scaling exponents and their associated complexity reflect that amplitude and phase play different roles in information processing and memory decoding : ( low - fidelity ) amplitude dynamics decode long - term memory , whereas the ( high - fidelity ) phase synchrony comprises a more complex and flexible memory coding ( in an information theoretic sense ) .
volume conduction can be a confounder in analyzing encephalographic recordings .
several methods to mitigate its effects have been proposed , generally relying on removing the instantaneous interactions that signify volume conduction @xcite . in consequence
these methods can only be applied in a pair - wise fashion , such that they are not applicable when considering the variables @xmath85 and @xmath86 .
we note , however , that volume conduction does not significantly influence the auto - correlation structure of the signals under study .
this finds support by shriki and coworkers @xcite who showed that mere linear mixing can not ` transform ' uncorrelated activity to power - law scaling .
in fact , if activity displays a power law , linear mixing does not alter this apart from slightly lowering the scaling exponent .
the occurrence of power laws is not only consistent with the criticality hypothesis @xcite .
the macroscopic behavior of self - organizing processes can | in general | be cast into a low - dimensional system when critical @xcite ; this in fact motivated looking at collective amplitude and phase synchrony .
several modeling studies support the seminal role of self - organization in neural dynamics and often highlight self - organized criticality .
for example , synaptic plasticity under the influence of a simple learning rule leads to scale - free networks and power - law distributions of avalanche dynamics @xcite . in resting state ,
the mechanism of local feedback mediated inhibition increases model performance @xcite and may also be interpreted as a form of self organization .
furthermore , the slow evolution of the order parameters in self - organizing systems is consistent with the time scale on which rsns evolve .
while previous work has shown that rsns fluctuate at slow ( @xmath87sec ) time scales @xcite , it has been recently shown that rsns are also expressed in meg activity on faster ( @xmath88sec ) time scales @xcite . alongside the results in this paper
, this represents mounting evidence that rsns are expressed across a range of time scales , i.e. they are time scale invariant . in summary
we have shown the presence of persistent long - range correlations in the evolution of global brain dynamics , i.e. the auto - correlation function obeys a power - law with scaling exponents exceeding those corresponding to random processes without memory .
this adds further support to the hypothesis that the brain is in a ( permanently ) critical state .
the here - reported scaling exponents clearly discriminate amplitude and phase dynamics , suggesting their differential role in whole - brain information processing . | converging research suggests that the resting brain operates at the cusp of dynamic instability signified by scale - free temporal correlations .
we asked if the scaling properties of these correlations differ between amplitude and phase fluctuations , which may reflect different aspects of cortical functioning .
using source - reconstructed magneto - encephalographic signals , we found power - law scaling for the collective amplitude and for phase synchronization , both capturing whole - brain activity .
the temporal changes of the amplitude comprise slow , persistent memory processes , whereas phase synchronization exhibits less temporally structured and more complex correlations , indicating a fast and flexible coding .
this distinct temporal scaling supports the idea of different roles of amplitude and phase in cortical functioning .
` power laws , criticality , dfa , amplitude , phase ` |
probing visible and dust - obscured star formation and physical properties of galaxies from the local to the early universe , including the epoch of cosmic reionization at redshift @xmath15 , is key to understanding the formation and evolution of galaxies , roles of galaxies in reionization and their relation among diverse galaxy populations . to date , objects such as quasars , active galactic nuclei ( agns ) , their host galaxies , submm - detected starburst galaxies ( smgs ) , normal star - forming galaxies and ( ultra)luminous infrared galaxies ( ( u)lirgs ) have been detected at rest frame far - infrared ( fir ) wavelengths in the local to high redshift universe ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the co rotational transition lines have been often used as a tracer of molecular gas to probe both visible and obscured sfrs , properties of molecular gas clouds and dynamics of galaxies . however , at @xmath15 , detections of the co line have been limited to only the quasars / agns , their hosts and an smg ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , because highly excited co lines only seen in these objects are bright , while the co lines in @xmath15 normally star - forming galaxies are likely not as highly excited and thus too faint to detect ( e.g. , * ? ? ?
* ; * ? ? ?
although sfrs of @xmath15 normally star - forming galaxies such as ly@xmath1 emitters ( laes ) and lyman break galaxies ( and related properties such as ly@xmath1/uv luminosity function / density ) are also an important probe of reionization , their dust - obscured sfrs and other relevant physical properties have been still poorly constrained ( e.g. , * ? ? ?
* ; * ? ? ?
? * ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
alternatively , the ionized carbon 157.74 @xmath3 m [ cii ] line ( rest frame frequency 1900.54 ghz ) has been observed to be the strongest cooling line of interstellar medium ( ism ) in many galaxies ( e.g. , a few hundred to a few thousand times stronger than co(10 ) , co(21 ) , co(32 ) and co(43 ) lines in @xmath162 star - formation / agn - dominated galaxies , * ? ? ?
* ) , and its luminosity is as high as @xmath17 0.011% of total fir luminosity ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
? * ; * ? ? ?
* ; * ? ? ?
when redshifted to @xmath15 , the [ cii ] line is observable in millimeter wavelengths from the ground .
even so , to date , only a handful number of quasar / agn host galaxies and an smg have been detected in [ cii ] at @xmath15 ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , and there is no detection in normally star - forming galaxies in the reionization epoch @xcite even though it is thought to be the strongest line in fir regime in these galaxies .
however , the advent of the atacama large millimeter / submillimeter array ( alma ) has the potential to change this situation as it has an order of magnitude higher sensivity than any other current millimeter interferometers . even in its early science operation
, it has been argued that alma could potentially detect the [ cii ] lines of normally star - forming galaxies in the reionization epoch at @xmath15 in only a few hours and resolve them on a few kpc scales .
motivated by this , we conducted a [ cii ] line and fir continuum observation of such a galaxy with alma . among many @xmath15 star - forming galaxies found to date
, we chose to observe a @xmath18 lae that we had discovered , iok-1 @xcite , because it is one of the handful of spectroscopically confirmed highest - class redshift ( @xmath19 ) objects , and we have a wealth of ancillary data availaible for this galaxy that were analyzed in the context of the alma observation . for example , we have conducted iok-1 s photometry and spectroscopy at optical to mid - infrared wavelengths and derived physical properties such as ly@xmath1 and uv luminosities / sfrs , stellar mass and morphology at uv wavelengths @xcite . in the present paper , we investigate the properties of iok-1 in the [ cii ] line and fir continuum .
the paper is organized as follows . in 2 , we describe our alma observation of iok-1 . then , in 3 , we analyze the properties of iok-1 in fir continuum and [ cii ] line such as sed , effects of cosmic microwave background ( cmb ) on dust continuum , dust mass , infrared ( ir ) luminosity and dust - obscured sfr as well as the relation between sfr and [ cii ] luminosity of high redshift laes and [ cii ] to fir luminosity ratio , and compare these properties with those of other galaxy populations in the local to high redshift universe , and discuss its implication for galaxy formation and evolution .
we summarize and conclude our study in 4 . throughout
, we adapt a concordance cosmology with @xmath20 .
we observed iok-1 by targeting its [ cii ] line redshifted to @xmath0 measured from ly@xmath1 emission ( 1.26 mm , 238.76 ghz in the observer s frame or 157.74 @xmath3 m , 1900.54 ghz at rest frame ) and the underlying dust continuum emission .
we used the alma band 6 with the extended array consisting of 1924 12-m diameter antennas on 15 june , 15 and 27 july and 29 november 2012 during the cycle 0 period .
we set the two spectral windows ( spws ) in the lower sideband to the continuum at the lower frequency side of the [ cii ] line , and the third and fourth spws in the upper sideband to the redshifted [ cii ] line and the continuum at the higher frequency side , respectively , in order to cover the continua at the both sides of [ cii ] line .
specifically , we set widths of all the four spws to 2 ghz ( 1.875 ghz after discarding edge channels ) using the time domain mode ( tdm ) for dual polarization with central frequencies of 223.76 , 225.76 , 238.76 and 240.76 ghz .
the channel spacing of the tdm mode is 15.6 mhz , and this gives a spectral resolution of 40 km s@xmath4 .
we obtained 2.7 hours of on - source integration .
we carried out data reduction with the common astronomy software applications ( casa ) .
the flux density was scaled with the observation data of 3c279 ( j1256 - 057 ) for one of the four execution blocks and titan for the rest of three execution blocks .
the typical flux uncertainties in the millimeter regime are @xmath21% .
j1310 + 323 was used for phase calibration .
3c279 was also used for bandpass calibration . finally , we imaged the [ cii ] line and fir continuum , using ` natural ' weighting , and reaching the sensitivities of @xmath2 215 @xmath3jy beam@xmath4 per tdm channel ( 40 km s@xmath4 ) and @xmath5 18 @xmath3jy beam@xmath4 for 7.5 ghz bandwidth ( 4 spws @xmath22 1.875 ghz ) , respectively .
the resolutions ( synthesized beam sizes ) of [ cii ] line and fir continuum images presented here are @xmath23 .
figure [ fir_continuum_iok1 ] shows the fir continuum map around iok-1 at @xmath23 resolution . the galaxy is not significantly detected .
iok-1 has a size of @xmath24 in the subaru telescope ly@xmath1
narrowband nb973 image @xcite , which is larger than the sizes in any other images ( rest frame uv continuum ) where iok-1 is detected .
hence , we convolved the fir continuum map to the resolution of @xmath25 ( through tapering with a 100 k@xmath26 taper ) to see if we detect any extended source
. however , we did not detect iok-1 in this map , either . as the rms of this map is @xmath27 @xmath3jy beam@xmath4 ,
the @xmath28 upper limit on 1.26 mm fir continuum of iok-1 is @xmath29 @xmath3jy . in this section ,
based on this flux limit , we constrain the dust properties of iok-1 such as its possible sed shape , dust mass , total ir luminosity and dust - obscured sfr by considering the effects of cosmic microwave background ( cmb ) on the observed continuum flux .
the study by @xcite describes how the cmb affects measurements of fir continuum emission of high redshift galaxies .
hence , before we derive and discuss constraints on the dust properties of iok-1 from the observed 1.26 mm continuum flux limit , we first describe the effects of cmb on fir continuum and how we actually apply them to our observed continuum flux limit ( see also * ? ? ? * for details ) .
cmb affects measurements of fir continuum in two ways .
one is that cmb heats the dust of galaxies to higher temperature as redshift gets higher .
according to @xcite , the dust temperature at a redshift @xmath30 is @xmath31)^{\frac{1}{4+\beta } } \label{tdust}\ ] ] here , @xmath32 is the temperature of dust heated by and in thermal equilibrium with the radiation field produced by stars in a galaxy at @xmath33 , where the effects of cmb radiation are negligible . also , @xmath34 k and @xmath35 are temperature of cmb at @xmath33 and dust emissivity index of a thermal spectrum .
this equation suggests that effect of cmb heating on dust temperature becomes non - negligible at high redshifts ( e.g. , see figure 1 in * ? ? ?
* for a case of @xmath36 k and @xmath37 ) .
the ratio ( hereafter referred to as @xmath38 ) between the dust continuum flux densities at a rest frame frequency @xmath39 and a redshift @xmath30 when including dust heating by cmb and not including it is @xmath40 where @xmath41 , @xmath42 and @xmath43 are the intrinsic dust continuum flux density of a galaxy , the dust continuum flux density after including the effect of cmb heating of dust and the planck function .
this @xmath38 is a factor for correcting the effect of cmb heating of dust ( for example , see figure 3 top panel of * ? ? ?
* for the case of @xmath36 k and @xmath44 ) .
meanwhile , another effect of cmb on dust continuum measurements is that cmb itself becomes a background against which fir continuum flux should be measured . @xcite
have shown that the ratio ( hereafter referred to as @xmath45 ) between the flux at a frequency @xmath39 that can be measured against cmb at a redshift @xmath30 and the flux emitted by dust in a galaxy at that frequency and redshift including heating of dust by cmb is @xmath46 where @xmath47 is a temperature of cmb at a redshift @xmath30 .
this is a factor for correcting the effect of cmb as background ( for example , see figure 3 bottom panel of * ? ? ?
* for the case of @xmath36 k and @xmath44 ) .
finally , we can estimate the intrinsic fir continuum flux @xmath48 emitted by a galaxy from the observed continuum flux @xmath49 by using the equation @xmath50 f_{\nu/(1+z)}^{\rm int } \label{f_int}\end{aligned}\ ] ] and the equation ( 1 ) by assuming @xmath32 and @xmath35 .
cccccccccccc 6.96 & 238.76 & 21 & 240 & @xmath51 & @xmath52 & @xmath53 & @xmath54 & @xmath55 & @xmath56 & @xmath57 & 23.9 since we have obtained @xmath28 upper limit on the observed 1.26 mm continuum flux of iok-1 , @xmath29 @xmath3jy , we can infer what type of galaxy this object is from its sed at rest frame uv to fir ( observer frame optical to millimeter ) including our alma flux limit . in figure [ sed_iok-1 ] ,
we compare the sed of iok-1 with template seds of several different types of local galaxies .
we reproduce the sed of iok-1 using the fluxes measured in the optical to fir images taken with subaru telescope suprime - cam @xmath58 bands @xcite , _ hubble space telescope _
( _ hst _ ) wide field camera 3 ( wfc3 ) f110w , f125w and f160w bands @xcite , _ spitzer space telescope
_ infrared array camera ( irac ) 3.6 , 4.5 , 5.8 and 8.0 @xmath3 m bands @xcite , _ herschel space telescope _
spectral and photometric imaging receiver ( spire ) 250 , 350 and 500 @xmath3 m bands ( the images of iok-1 are cut out from the completed _ herschel_-atlas north galactic plane ( ngp ) project images ; * ? ? ?
* ; * ? ? ?
* ivison , eales & smith 2013 private communication ) and our alma @xmath28 upper limit on 1.26 mm dust continuum .
these images are shown in figure [ images_iok1 ] .
iok-1 is not detected in any of the spire bands where the beam - convolved map @xmath59 point source sensitivities ( including both instrumental and confusion noises ) are 4.6 , 5.3 , and 6.0 mjy for 250 , 350 and 500 @xmath3 m bands , respectively .
note that for comparison , in figure [ sed_iok-1 ] , we also plot @xmath28 upper limits on 1.26 mm continuum from the previous iram / plateau de bure interferometer ( pdbi ) and combined array for research in millimeter - wave astronomy ( carma ) observations of iok-1 conducted by @xcite and @xcite , which are shown by the lower and upper open triangles , respectively .
our alma upper limit is about one order of magnitude deeper than these previous limits .
we compare the uv fir fluxes / flux limits of iok-1 with the local galaxy templates of arp 220 , m82 , m51 , ngc 6946 ( the left panel in figure [ sed_iok-1 ] ) taken from @xcite and other local spiral / barred , dwarf and irregular galaxies ( mrk 33 , ngc 0925 , ic 2574 , holmberg ii and ddo 165 in the left panel and all the galaxies in the right panel in figure [ sed_iok-1 ] ) from @xcite .
we shift these galaxy seds to the redshift of iok-1 , @xmath0 , normalize them to the rest frame uv continuum flux density redward of ly@xmath1 ( the _ hst_/f125w band flux ) and plot them by the color coded solid lines in figure [ sed_iok-1 ] .
we also show their galaxy morphological types in the figure .
these are the seds not including the effects of cmb on their fir continua .
we also plot the seds of these template galaxies with their fir continua ( a part of the sed that corresponds to the emission of the cold dust ) including the effects of cmb by the color coded dashed lines in figure [ sed_iok-1 ] .
we convert the original fir continua of the template galaxies , @xmath48 , to those including the effects of cmb , @xmath49 , by using their dust temperatures @xmath32 and emissivity indices @xmath35 obtained from literature and equations ( [ tdust])([f_int ] ) but following the prescription given by @xcite to calculate the factor @xmath38 as follows .
first , from the literature that conducted single - component modified black body ( mbb ) fittings to the observed fir continua of the galaxies , we adopt @xmath60 k and @xmath61 for arp 220 @xcite , @xmath62 k and @xmath63 for m82 @xcite , @xmath64 k and @xmath37 for m51 @xcite , @xmath65 k and @xmath66 for mrk 33 @xcite , @xmath67 k and @xmath68 for ngc 6946 , @xmath69 k and @xmath68 for ngc 0925 , @xmath70 k and @xmath68 for ic 2574 , @xmath71 k and @xmath68 for holmberg ii , @xmath72 k and @xmath68 for ddo 165 @xcite .
these are the galaxies shown in the left panel of figure [ sed_iok-1 ] .
we adopted @xmath32 s and @xmath35 s obtained by @xcite for all the galaxies shown in the right panel of figure [ sed_iok-1 ] .
then , to compute the factor @xmath38 at each frequency @xmath39 for each template galaxy , we scale the mbb with @xmath32 and @xmath35 to the peak of the fir emission of the galaxy s sed and calculate the ratio @xmath73 of emission associated with the cold dust at each frequency @xmath39 .
@xmath74 where @xmath75 is the scaling factor . with this @xmath73 , the equation ( [ m_nu ]
) can be now expressed as @xmath76 using the equations ( [ tdust ] ) and ( [ c_nu])([m_nu2 ] ) , we convert the original fir continuum emission of each template galaxy at each frequency @xmath39 to the one including the effects of cmb and show the results in figure [ sed_iok-1 ] .
comparison of the solid lines ( _ intrinsic _ dust seds ) and the dashed lines ( _ observed _ dust seds which include the effects of cmb ) in figure [ sed_iok-1 ] shows two notable differences .
first , due to the extra heating of dust by cmb at @xmath0 , the peaks of the seds shift toward shorter wavelengths ( or higher frequencies ) . also , the rayleigh - jeans tails of the dust emission observed against the cmb ( dashed lines ) are steeper than those of the intrinsically emitted seds ( solid lines ) .
this is because at a higher redshift cmb temperature is closer to the dust temperature , making the intrinsic dust emission more difficult to detect against cmb .
the effects of cmb are stronger for galaxies with lower dust temperatures @xmath32 .
especially , as the cmb temperature is @xmath77 k at @xmath78 , seds of the template galaxies with @xmath32 closer to this ( i.e. , m51 , ngc 6946 , ngc 0925 , ic 2574 and ddo 165 ) are more remarkably affected by the cmb effects than those with @xmath32 much higher than 21.84 k. as discussed in @xcite , considering the effects of the cmb is required for the correct interpretation of dust continuum observations as it would change estimation of a dust continuum flux density and thus subsequent derivations of dust masses , total ( f)ir luminosities and dust - obscured sfrs of galaxies .
figure 2 also shows that the sed of iok-1 is neither similar to those of highly dust - obscured galaxies such as arp 220 and m82 ( see also * ? ? ?
* ; * ? ? ?
* ) nor less extreme dust - rich nearby spirals such as m51 and ngc 6946 .
it is rather more likely similar to seds of dwarfs / irregulars .
more specifically , uv continuum slopes ( see the insets in figure [ sed_iok-1 ] ) and fir continuum seds of ic 2574 ( morphological type sabm ) , holmberg ii ( i m ) , ddo 053 ( i m ) , holmberg i ( iabm ) and ngc 2915 ( i0 ) are consistent with the detection data points of iok-1 at rest frame uv and the alma fir continuum upper limit ( morphological types were taken from * ? ? ?
as local dwarf / irregular galaxies have been observed to have low ir luminosities and dust masses , iok-1 could be such a system with low dust obscuration .
meanwhile , conducting sed - fitting to rest frame uv to optical fluxes , @xcite estimated the upper limit on the stellar mass of iok-1 to be either @xmath79@xmath80 for a young age ( @xmath81 myr ) and low dust reddening ( @xmath82 ) or a mass as high as @xmath83@xmath84 for either an old age ( @xmath85 myr ) or high dust reddening ( @xmath86 ) .
given that laes at high redshifts ( e.g. , @xmath876.6 ) tend to be very young and low dust systems ( e.g. , * ? ? ?
* ) , and considering the non - detection of iok-1 in dust continuum with alma , the former @xmath88 limit would be more likely . for comparison , we have calculated the average stellar mass of 17 local dwarf and irregular galaxies ( sd and later ) in @xcite kingfish survey sample and obtained @xmath89 ( the stellar mass ranges from @xmath90 to @xmath91 within the sample ) .
hence , the stellar mass of iok-1 is comparable to those of local dwarf / irregular galaxies .
this also supports the idea that iok-1 would be a system similar to local dwarf / irregular galaxies .
given our @xmath28 alma upper limit on 1.26 mm continuum flux density @xmath92 , we can estimate the upper limit on the dust mass @xmath93 of iok-1 from the following relation .
@xmath94 } \nonumber \\ \label{dustmass}\end{aligned}\ ] ] here , @xmath95 is a luminosity distance , @xmath96 is the rest frame frequency dust mass absorption coefficient , and @xmath97 is the planck function at a rest frame frequency @xmath98 and a dust tempatature @xmath99 . we adopt @xmath100 @xmath101 g@xmath4 from dunne et al .
( 2000 ) with the rest frame wavelength of @xmath102 @xmath3 m .
@xmath48 is the intrinsic continuum flux that can be obtained after correcting the observed continuum flux @xmath103 @xmath3jy for the cmb effects by using the equation ( [ f_int ] ) . because the sed of iok-1 is similar to those of local dwarf and irregular galaxies ( see [ sed ] and figure [ sed_iok-1 ] ) , we model the dust continuum sed of iok-1 as an mbb with @xmath104 k and @xmath68 , the average values of 17 local dwarf and irregular galaxies ( sd and later ) derived by @xcite from their kingfish survey
then , the equation ( [ dustmass ] ) gives @xmath105 .
the upper limit on dust mass estimated here is also listed in table 1 .
if we do not correct the effects of cmb ( i.e. , if we use the observed continuum flux limit @xmath103 @xmath3jy with @xmath106 instead of @xmath107 for the equation ( [ dustmass ] ) ) , this leads to @xmath108 , a factor of 0.76 ( @xmath109 ) underestimation compared to the intrinsic dust mass limit . meanwhile ,
if @xmath32 of iok-1 was actually lower ( higher ) than what we assumed here ( 27.6 k ) , the upper limit on @xmath93 would be higher ( lower ) .
we also estimate the upper limit on intrinsic fir luminosity @xmath14 of iok-1 by again modeling its dust continuum sed as an mbb with @xmath104 k and @xmath68 ( the average values of local dwarfs and irregulars derived by * ? ? ?
* see above ) , scaling it to our cmb - corrected @xmath28 alma upper limit on 1.26 mm continuum ( i.e. , @xmath110 ) and integrating it from 42.5 @xmath3 m to 122.5 @xmath3 m .
we obtain @xmath111 . moreover , integrating the same scaled mbb from 8 @xmath3 m to 1 mm , we estimate the intrinsic total ir luminosity to be @xmath9 .
these intrinsic luminosities correspond to only the contribution from the stars in iok-1 to dust heating and do not include the contribution from cmb .
hence , especially , @xmath112 is appropriate for estimating the dust - obscured sfr .
if we use the @xcite relation with the @xcite initial mass function ( imf ) , sfr@xmath113 @xmath114 $ ] , the total ir luminosity converts to the dust - obscured sfr of sfr@xmath115 @xmath11yr@xmath4 .
@xcite estimated the sfr of iok-1 to be sfr@xmath116 @xmath11yr@xmath4 from the hst / wfc3 band magnitudes which cover the rest frame uv continuum of iok-1 not including ly@xmath1 emission .
this is the sfr not obscured by dust .
hence , the upper limit on the total ( obscured plus un - obscured ) sfr of iok-1 is sfr@xmath117 sfr@xmath118 sfr@xmath119 @xmath11yr@xmath4 .
this implies that @xmath120 sfr@xmath121sfr@xmath122% of star formation is obscured by dust in iok-1 . on the other hand ,
if we scale the mbb to just the observed continuum flux density limit @xmath49 ( i.e. , the flux not corrected for the cmb effects ) and integrate it , this would result in underestimations of @xmath14 , @xmath112 and sfr@xmath123 by 24% ( a factor of @xmath124 ) compared to the intrinsic ones we have obtained above , due to the extra heating of dust by cmb and the reduction of observed dust continuum flux by the cmb as a background .
the dust mass and fir / ir luminosity of iok-1 are @xmath1252 orders of magnitude lower than those of a host galaxy of the quasar ulas j112001.48 + 064124.3 at a similar redshift @xmath126 @xcite and the most distant smg known , hfls 3 , at @xmath127 @xcite as well as those of the quasar host galaxies at @xmath1286.4 studied so far @xcite and @xmath1295 smgs @xcite . furthermore ,
though the sensitivities are about one order of magnitude lower than that of alma , @xcite also observed iok-1 in dust continuum , @xcite and @xcite carried out 1.2 mm continuum observations of three other @xmath130 laes , and they all ended up with non - detections . our and their results together imply that high redshift laes at the epoch of reionization are likely systems with very little dust , quite different in dust content from extreme starbursts or highly dust - obscured galaxies such as quasar hosts and smgs at similar and lower redshifts . moreover , @xcite observed dust continuum of a @xmath131 lae , himiko , which has a similar sfr@xmath132 ( @xmath133 @xmath11yr@xmath4 ) to iok-1 , with alma to a similar sensitivity to ours at @xmath134 resolution but found only a @xmath135 weak signal at the expected position and considered it a non - detection .
thus , himiko could be similar to typical high redshift laes in dust content .
however , himiko has at least twice larger size than iok-1 in ly@xmath1 ( @xmath136 in extent ) and is considered a ly@xmath1 blob ( lab ) .
also , its uv continuum shows a few knots over about @xmath137 @xcite .
hence , to search for possible extended emission , we have re - calibrated , flagged , and imaged the public himiko alma cycle 0 data .
we made the continuum map that is based on all data except the spw in which the [ cii ] line is expected , to avoid a possible contamination from that line .
as seen in figure [ himiko ] , at a nearly full resolution of @xmath138 , we found that there is a @xmath139 possible weak continuum source at the position of himiko , as pointed out by @xcite .
the source shows a possible extension east - west , along the uv continuum direction , although low signal to noise ratio does not allow us to confirm if the source is extended .
if we consider this @xmath139 signal the upper limit ( as it is marginal ) and model the dust sed of himiko as an mbb with @xmath104 k and @xmath68 typical of dwarf / irregular galaxies , integrations of the mbb scaled to 67.5 @xmath3jy/@xmath140 ( @xmath141 for himiko ) give @xmath142 and @xmath143 .
the @xcite relation converts the @xmath112 to sfr@xmath144 @xmath11yr@xmath4 and sfr@xmath145 @xmath11yr@xmath4 , implying that @xmath146% of star formation is obscured by dust .
also , the equation ( [ dustmass ] ) gives dust mass of himiko , @xmath147 @xmath11 .
meanwhile , imaging at @xmath148 resolution ( comparable to the ly@xmath1 size of himiko ) resulted in a higher noise , and a non - detection with an rms of 42 @xmath3jy beam@xmath4 .
integrations of the same mbb scaled to the @xmath28 limit give the conservative upper limits , @xmath149 , @xmath150 , sfr@xmath151 @xmath11yr@xmath4 , sfr@xmath152 @xmath11yr@xmath4 and @xmath153% .
also , the equation ( [ dustmass ] ) gives the conservative dust mass upper limit , @xmath154 @xmath11 .
hence , although it is not clear that himiko is extended in dust continuum , it could be similar to other common laes at high redshifts such as iok-1 in dust content or possibly slightly dustier if the @xmath139 signal is a real detection .
it should be noted that comparison of these continuum flux limits at @xmath138 and @xmath148 resolutions with dust seds of different types of local galaxies in figure 5 in @xcite implies that himiko could be a system similar to local dwarf / irregular galaxies , not changing the same implication obtained by @xcite by using their derived himiko continuum limit .
thus , our assumption of the mbb with @xmath32 and @xmath35 typical of local dwarf / irregular galaxies for himiko to derive its dust - related properties is reasonable .
figures [ cii_iok1image ] and [ cii_iok1spec ] show the [ cii ] channel maps and the spectrum at a resolution of @xmath23 around the expected redshift of iok-1 based on ly@xmath1 emission . we see a @xmath6 weak signal at the expected position of iok-1 with a peak at @xmath155 ghz ( the peak flux density is @xmath156 @xmath3jy ) and an fwhm of @xmath157 ghz or @xmath158 km s@xmath4 ( there are also similar @xmath28 features at 239.05 ghz and 238.02 ghz , and we will mention these below ) . if the signal at @xmath155 ghz was indeed the [ cii ] line , the redshift based on its peak is @xmath159}=6.9585 $ ] .
it is slightly lower than the redshift measured from ly@xmath1 emission peak : @xmath160 ( * ? ? ?
* using subaru telescope focas spectrograph ) ; @xmath161 ( * ? ? ? * using keck telescope deimos spectrograph with a slightly better spectral resolution ) by @xmath162245 km s@xmath4 .
this is consistent with the idea that @xmath163 could be overestimated due to the absorption of blue side of ly@xmath1 emission by the intergalactic medium .
also , previous spectroscopic studies of laes and lbgs at @xmath1293 have shown that their ly@xmath1 emission peaks are usually shifted redward of the redshifts accurately measured from other lines and absorptions ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
iok-1 is not only an lae detected in a narrowband @xcite but also an lbg detected as a @xmath164-band dropout @xcite .
thus , if the @xmath28 feature is indeed the [ cii ] line , it is also consistent with the trend that redshifts of laes and lbgs measured from ly@xmath1 tend to be higher than their systemic redshifts .
however , the @xmath28 significance of this signal is marginal and is not strong enough to definitively conclude that the detected feature is indeed the [ cii ] line .
hence , in this paper , we consider it a non - detection and leave the firm conclusion about detection / non - detection to our forthcoming ( scheduled ) observations of iok-1 with alma . note that there are also similar @xmath28 features at 239.05 ghz and 238.02 ghz .
however , as for the one at 239.05 ghz , this portion of the spectrum is noisier than the one at frequencies @xmath165 ghz due to atmospheric absorption .
the feature at 238.02 ghz is unlikely to be [ cii ] line , because if it was [ cii ] , the redshift derived from it would be higher than the redshift estimated from ly@xmath1 emission .
as we mentioned earlier , iok-1 has a size of @xmath24 in ly@xmath1 , larger than the beamsize of the [ cii ] map @xmath23 .
hence , we also made the [ cii ] map convolved to a @xmath25 resolution ( through tapering with a 100 k@xmath26 taper ) comparable to the ly@xmath1 size to see if we detect any extended source .
however , we did not detect iok-1 in this map , either . as the rms of this map is @xmath166 @xmath3jy beam@xmath4 ,
the @xmath28 upper limit on [ cii ] line flux density of iok-1 is 720 @xmath3jy beam@xmath4 over a channel width of 40 km s@xmath4 ( same as the highest spectral resolution at our observing frequency of 238.76 ghz in the tdm mode we used ) .
this converts to a line luminosity of @xmath13 } < 3.4 \times 10 ^ 7 l_{\odot}$ ] .
we assumed a line width of @xmath167 km s@xmath4 .
@xcite simulated the [ cii ] line of iok-1 by using the same procedure as the one of @xcite who combined high resolution , radiative transfer cosmological simulations of @xmath168 galaxies with a sub - grid multi - phase model of their interstellar medium and derived the expected intensity of several fir emission lines including [ cii ] for different values of the gas metallicity . conservatively assuming solar metallicity
, their simulation suggests that most of the [ cii ] emission of iok-1 comes from a cold neutral medium and that fwhm of its main peak is @xmath169 km s@xmath4 . in this context
, our assumption of @xmath167 km s@xmath4 for calculating the [ cii ] line luminosity upper limit is reasonable . in this section ,
based on this limit , we explore the properties of iok-1 related to the [ cii ] line .
note that the effects of cmb on [ cii ] line observations of high redshift galaxies were discussed and found to be negligible by @xcite .
hence , we do not consider the effects of cmb on [ cii ] line in the following .
the [ cii ] line has been considered a tracer of sfr as it is produced by the uv field in star forming regions .
figure [ sfr_vs_lcii ] shows the [ cii ] luminosity as a function of sfr for several different types of objects .
objects with higher sfrs have higher [ cii ] luminosities , and different types of objects reside in different locations in the sfr@xmath13}$ ] plane in such a way that local late - type and star - forming galaxies are located at the low to intermediate sfr and @xmath13}$ ] regime , low redshift lirgs at the intermediate to high sfr and @xmath13}$ ] region and quasar host galaxies and smgs at the very high sfr and @xmath13}$ ] domain . for the local late - type and star - forming galaxies , clear correlations have been observed to hold between sfr and @xmath13}$ ] @xcite . in figure
[ sfr_vs_lcii ] , we compare these calibrated relations for local galaxies and locations of all types of objects in the sfr@xmath13}$ ] plane with total sfrs and [ cii ] luminosities of iok-1 as well as other laes at @xmath1706.6 previously observed to see whether they are similar or different systems in gas enrichment or in metallicity as [ cii ] luminosity would be proportional to metallicity ( e.g. , * ? ? ? * ) . in [ fir_properties ] , we have estimated an upper limit on the total sfr of iok-1 to be sfr@xmath171 @xmath11yr@xmath4 by adding the sfr estimated from its uv continuum ( sfr@xmath132 ) and the upper limit on the dust - obscured sfr estimated from 1.26 mm continuum ( sfr@xmath123 ) .
if iok-1 had no dust , the lowest possible total sfr would be the one estimated from its uv continuum , i.e. , sfr@xmath172 sfr@xmath173 @xmath11yr@xmath4 @xcite . in figure [ sfr_vs_lcii ]
, we also plot the possible total sfr range of iok-1 and the upper limit on [ cii ] luminosity .
we notice that iok-1 neither seems to follow the sfr@xmath13}$ ] correlations for local star - forming galaxies ( though the correlation has some scatters ) nor to be within the regions occupied by any other type of objects except for high redshift laes ( i.e. , himiko and hcm 6a which we will mention below ) .
its [ cii ] luminosity is significantly fainter than that expected for a local star - forming galaxy having the same sfr as iok-1 . by using the local star - forming galaxy correlation derived by @xcite , the [ cii ] luminosity limit of iok-1 converts to sfr of @xmath174 @xmath11yr@xmath4 ( or @xmath175 sfr @xmath176 ) , which is much lower than the possible sfr range of iok-1 , 23.933.8 @xmath11yr@xmath4 .
hence , the local star - forming galaxy correlation does not seem to apply to iok-1 .
iok-1 could be a system having either lower enriched gas or different photodissociation region ( pdr ) structure / physical state of ism than local star - forming galaxies .
a similar result has been reported for a @xmath131 lae , himiko , by @xcite who did not detect its [ cii ] line by using alma at @xmath134 resolution and 200 km s@xmath4 channel@xmath4 .
as himiko is extended over @xmath136 in ly@xmath1 , we generated the spectrum at @xmath177 resolution and 30 mhz channel@xmath4 ( 36 km s@xmath4 ) , using the public himiko alma cycle 0 data , to see if we detect any extended source .
as with the higher resolution analysis of @xcite , no line is detected .
the 3@xmath178 flux density upper limit at @xmath177 resolution is 2.0 mjy beam@xmath4 channel@xmath4 at 36 km s@xmath4 channel@xmath4 or @xmath13 } < 7.9 \times 10 ^ 7 l_{\odot}$ ] ( assuming a velocity width of @xmath179 km s@xmath4 ) , and a factor 1.7 lower at 100 km s@xmath4 channel@xmath4 .
however , we have found that the rms noise is about a factor two higher than this between 249.7 ghz to 250.2 ghz due to atmospheric absorption , so if the line happened to fall in this frequency range , the limits are a factor of two worse . in [ fir_properties ] , we also derived the conservative upper limit on sfr@xmath180 of himiko at @xmath148 resolution ( comparable to the ly@xmath1 size of himiko ) . in figure
[ sfr_vs_lcii ] , we plot the @xmath13}$ ] upper limit and the possible sfr range of himiko ( sfr@xmath181 @xmath11yr@xmath182 sfr@xmath183 sfr@xmath132 @xmath184 sfr@xmath185 @xmath11yr@xmath4 ) .
himiko is off the local galaxy sfr@xmath13}$ ] correlation and close to iok-1 . for comparison , in figure [ sfr_vs_lcii ] , we also plot the @xmath13}$ ] upper limit of himiko we derived with sfr@xmath186 @xmath11yr@xmath4 which @xcite estimated from sed - fitting at rest frame uv to optical wavelengths . in this case
, himiko is also off the local galaxy sfr@xmath13}$ ] correlation but apart from iok-1 due to the difference in sfr@xmath180 .
both iok-1 and himiko have several things in common except that himiko displays more extended ly@xmath1 emission and is considered an lab and that himiko might be detected in dust continuum .
they both consist of 23 merging components @xcite , have similar total uv continuum luminosities ( or sfr@xmath132 ) of the whole system and have not been detected in [ cii ] line to the similar flux limits ( if we assume that the sizes of these laes in [ cii ] are similar ) .
hence , if sfr@xmath132 @xmath184 sfr@xmath123 more likely reflects the actual total sfr of himiko than the sfr estimated from the sed - fitting , both iok-1 and himiko may be normally star - forming galaxies at high redshifts with lower enriched gas / metallicity content than local normal star - forming galaxies . on the other hand
, @xcite observed a gravitationally lensed @xmath187 faint lae , hcm 6a , in [ cii ] emission with the iram / pdbi to a deep flux limit but somewhat shallower than ours and ouchi et al.s taking advantage of the lensing magnification .
they could not detect [ cii ] line , either . given sfr@xmath188
@xmath11yr@xmath4 @xcite , sfr@xmath189 @xmath11yr@xmath4 ( we derived this sfr in the same way as we did for iok-1 and himiko for a fair comparison ) and the @xmath13}$ ] upper limit ( * ? ? ?
* they adopted 100 km s@xmath4 width ) , we also plot hcm 6a in figure [ sfr_vs_lcii ] . it is close to the local sfr@xmath13}$ ]
relation although it could be off the relation as the [ cii ] luminosity is an upper limit .
this is possibly due to their slightly shallower [ cii ] flux limit .
since this lae is a factor of 23 lower than iok-1 and himiko in sfr@xmath132 , its [ cii ] luminosity could also potentially be lower than those of iok-1 and himiko . in figure [ sfr_vs_lcii ]
, we also plot the @xmath190 and @xmath191 laes , sdf j132408.3 + 271543 and sdf j132415.7 + 273058 @xcite , whose [ cii ] lines and 1.2 mm continua were observed by @xcite with carma and undetected .
we plot the sfr ranges estimated from sfr@xmath132 s and the sfr@xmath123 upper limits ( corrected for the cmb effects on their 1.2 mm continua ) derived by @xcite and @xcite , respectively .
they are within the local sfr-@xmath13}$ ] relation and the low redshift lirg region , but the @xmath13}$ ] upper limits ( 50 km s@xmath4 channel width ) obtained by @xcite are an order of magnitude shallower than those for iok-1 , himiko and hcm 6a .
therefore , at this moment , we can neither tell whether these two laes follow the local relation nor whether these results are consistent with those of iok-1 , himiko and hcm 6a .
deeper observations are required to reveal this . on the contrary , a @xmath170 lae , ly@xmath1 - 1 , observed with alma , is detected in [ cii ] line @xcite and located in the local sfr@xmath13}$ ] relation as well as close to the lower sfr and lower @xmath13}$ ] side of the low-@xmath30 lirgs region and the border between data points of the lirgs and the local star - forming galaxies , as seen in figure [ sfr_vs_lcii ] .
we plot the sfr range estimated from the sfr@xmath132 and the sfr@xmath123 upper limit derived by @xcite and @xcite , respectively .
ly@xmath1 - 1 would be a system different from @xmath1307 laes and possibly similar to low-@xmath30 lirgs .
actually , it is in the vicinity of a system br 1202 - 0725 comprising a quasar , an smg and another lae , ly@xmath1 - 2 , all at @xmath170 ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
note that we do not plot ly@xmath1 - 2 in figure [ sfr_vs_lcii ] because , though detected by alma , its @xmath13}$ ] can not be measured accurately due to the partial truncation at the edge of the alma band @xcite .
though upper limit , ly@xmath1 - 1 has an fir luminosity @xmath192 @xcite , comparable to those of lirgs ( @xmath193 ) . in addition , ly@xmath1 - 2 has @xmath194 @xcite , even as luminous as ulirgs ( @xmath195 ) . moreover , both laes have ly@xmath1 emission broader ( @xmath196 km s@xmath4 ) than most laes at similar redshifts @xcite , and ly@xmath1 - 1 is also spatially extended in the ly@xmath1 narrowband image @xcite .
several studies have reported that laes at @xmath1294.5 detected in fir continua and having @xmath14 as luminous as ulirgs also exhibit extended ly@xmath1 emission , and some of them are even classified as labs showing evidence of interaction within the ly@xmath1 cloud .
they also tend to be found in dense environments such as protoclusters @xcite . detecting new faint sources , possibly additional laes
also associated with the quasar and the smg , @xcite argue that the br 1202 - 0725 system might be a massive protocluster in which the high star formation activity is probably triggered by minor mergers or interactions .
hence , ly@xmath1 - 1 is probably a system not like usual laes such as iok-1 but rather possibly a fainter and less extended analog of lirgs / labs seen in dense environments .
also , at @xmath170 , the universe is a factor @xmath1971.7 older than the one at @xmath1987 .
hence , the difference between ly@xmath1 - 1 and the @xmath1987 laes may partly originate from galaxy evolution over cosmic time as well . in figure [ sfr_vs_lcii ] , we also plot a @xmath199 triple lbg system , lbg-1 , detected in [ cii ] but undetected in dust continuum with alma and associated with a @xmath200 protocluster including a @xmath201 smg , aztec-3 @xcite .
lbg-1 has an sfr comparable to those of @xmath2027 laes in figure [ sfr_vs_lcii ] , but its @xmath13}$ ] is about an order of magnitude higher than the laes ( except for sdf j132408.3 + 271543 and sdf j132415.7 + 273058 due to their shallow @xmath13}$ ] limits ) .
moreover , lbg-1 is on the two local galaxy sfr@xmath13}$ ] correlations and at the high sfr and high @xmath13}$ ] edge of the local star - forming galaxy data distribution . according to the analysis of @xcite ,
though lbg-1 comprises three lbg components and is associated with a protocluster , it does not seem to be a galaxy of extreme type but rather `` typical '' , @xmath203 star - forming system with relatively low dust content .
furthermore , @xcite compared the sed of lbg-1 including their alma dust continuum upper limit with those of local galaxies ( mostly the same local galaxy templates we and @xcite used for the same purpose for iok-1 and himiko ) and found that lbg-1 is similar to local dwarf / irregular galaxies .
hence , all the information taken together implies that both typical laes and lbgs would share seds similar to those of local dwarf / irregular galaxies , but lbgs would be more enriched in gas ( or higher metallicity ) or have different pdr structure / ism physical state than laes .
however , to securely confirm this , more observations and sample statistics of typical high redshift lbgs and laes are required . also , note that lbg-1 is not detected in dust continuum , but the alma observaton of this object by @xcite is a factor 37 shallower than those of iok-1 , himiko and hcm 6a .
thus , observations of lbgs and laes to comparable depth is necessary to fairly compare these two galaxy populations in the context of dust content .
figure [ lciioverlfir_vs_lfir ] shows the luminosity ratio @xmath13}/l_{\rm fir}$ ] as a function of fir luminosity @xmath14 plotted for different types of objects : @xmath204 lirgs @xcite , local normal galaxies @xcite , @xmath2052 galaxies including starburst - dominated , agn - dominated and mixed systems @xcite , @xmath2067.1 quasar host galaxies @xcite , @xmath2076.34 submm galaxies @xcite and a well known dwarf irregular galaxy , large magellanic cloud ( lmc ) @xcite as well as @xmath1707 laes , iok-1 ( this study ) , himiko @xcite , hcm 6a @xcite , sdf j132408.3 + 271543 , sdf j132415.7 + 273058 @xcite ,
ly@xmath1 - 1 in the quasar @xmath184 smg system bri 1202 - 0725 @xcite and a @xmath199 triple lbg system lbg-1 associated with a @xmath200 protocluster @xcite .
the [ cii ] line is the @xmath208p@xmath209 @xmath210 @xmath208p@xmath211 fine - structure line that could be emitted by the gas exposed to far - uv radiation in the pdrs , x - ray dominated regions ( xdrs ) related to an agn activity and/or hii regions ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
hence , the fraction of [ cii ] line luminosity compared to the total fir luminosity varies with type of object , reflecting its structure of pdr , xdr and/or hii region and physical state of ism .
for example , normal star - forming galaxies in the local universe tend to show very strong [ cii ] emission , typically nearly up to an order of magnitude stronger luminosity ratio ( @xmath212}/l_{\rm fir } \sim -2.5 $ ] on average ) than agn powered systems such as quasar host galaxies , ( u)lirgs and smgs .
hence , @xmath13}/l_{\rm fir}$ ] versus @xmath14 diagram can distinguish between different types of objects . in figure [ lciioverlfir_vs_lfir ] , we compare the @xmath1707 laes with other objects .
the color coded filled and open circles with arrows indicate their data with @xmath14 corrected and uncorrected for the cmb effects , respectively . for iok-1 , himiko and hcm 6a , we calculated their @xmath14 limits with and without the cmb effects corrected by integrating from 42.5 @xmath3 m to 122.5 @xmath3 m the mbbs with @xmath104 k and @xmath68 ( average of dwarf / irregular galaxies ; * ? ? ?
* ) scaled to the upper limits on the fir continuum flux density , which we corrected and did not correct for the cmb effects .
for himiko , we plot the data based on @xmath14 upper limits at @xmath213 resolution ( comparable to the size of himiko in ly@xmath1 ) where we assume that himiko is not detected in dust continuum at any resolution as well as the data based on the @xmath14 s at @xmath214 resolution where we assume that himiko is detected in dust continuum at 3.4 @xmath178 significance at this resolution ( see the text in [ fir_properties ] for details ) . for hcm 6a , we used its fir continuum flux density limit taken from @xcite . for sdf j132408.3 + 271543 and sdf j132415.7 + 273058
, we calculated their @xmath14 limits with and without the cmb effects corrected by integrating from 42.5 @xmath3 m to 122.5 @xmath3 m the template sed of ngc 6946 @xcite scaled to their upper limits on the fir continuum flux density provided in @xcite , as they have done so .
for each of @xmath2157 laes , we used its @xmath13}$ ] limit over a channel width we and each author adopted : 40 , 36 , 100 , 50 and 50 km s@xmath4 for iok-1 , himiko , hcm 6a , sdf j132408.3 + 271543 and sdf j132415.7 + 273058 , respectively @xcite .
we also plot the data of ly@xmath1 - 1 derived by @xcite with the asterisks .
the arrows of each lae data indicate the upper limits on the luminosity ratio @xmath13}/l_{\rm fir}$ ] at a given @xmath14 lower than the upper limit on the @xmath14 of that lae . comparing the filled and open circles in figure [ lciioverlfir_vs_lfir ] ,
we notice that omitting to correct for the cmb effects results in underestimation of @xmath14 and thus overestimation of the ratio @xmath13}/l_{\rm fir}$ ] as discussed in [ fir_properties ] .
as seen in figure [ lciioverlfir_vs_lfir ] , iok-1 , himiko and hcm 6a are in the region occupied by local normal star - forming galaxies
. however , these laes are apart from the average point of local normal star - forming galaxies , where @xmath216 and @xmath212}/l_{\rm fir } \sim -2.5 $ ] , and iok-1 and hcm 6a are even at the lower @xmath13}/l_{\rm fir}$ ] edge of the local normal star - forming galaxy region
. meanwhile , sdf j132408.3 + 271543 and sdf j132415.7 + 273058 are in the low redshift lirg region , probably because their upper limits on @xmath13}$ ] and @xmath14 are an order of magnitude shallower than those of other laes .
unless we observe them to much deeper limits , we can not tell if they are also located in the similar positions as the other @xmath1307 laes in figure [ lciioverlfir_vs_lfir ] .
we also notice that the possible location of iok-1 and hcm 6a in figure [ lciioverlfir_vs_lfir ] includes the data point of lmc , the most nearby dwarf irregular galaxy
. this might be consistent with a picture that high redshift laes could be similar to local dwarf / irregular galaxies , as inferred from the comparison of the sed of iok-1 with those of local dwarfs / irregulars in figure [ sed_iok-1 ] . however , this dose not necessary mean that the similarity in dust continuum seds guarantees the similarity in @xmath13}/l_{\rm fir}$ ] ratios , as we have not detected [ cii ] lines and dust continua of iok-1 and hcm 6a and uncertainties in their @xmath13}/l_{\rm fir}$ ] are large .
in fact , while the possible location of himiko in figure [ lciioverlfir_vs_lfir ] in the case that we assume a non - detection in dust continuum as an extended source ( the magenta filled circle ) includes the data point of lmc , the @xmath13}/l_{\rm fir}$ ] upper limit of himiko in the case that we assume a detection in dust continuum ( the black filled circle ) is not consistent with lmc , although both cases ( i.e. , the @xmath14 upper limit and `` detected '' @xmath14 value ) are consistent with the dust continuum seds of the dwarf / irregular galaxies in figure 5 of @xcite .
thus , it seems that the similarity in dust continuum sed does not necessarily mean similar @xmath13}/l_{\rm fir}$ ] ratios .
also , we neither have the dust continuum sed data of lmc that can be compared with those of iok-1 and himiko nor @xmath13}/l_{\rm fir}$ ] ratio data of dwarf / irregular galaxies whose dust continuum seds were compared with those of iok-1 and himiko in figure [ sed_iok-1 ] in this paper and figure 5 in @xcite .
thus , at this moment we can not tell if high redshift laes are similar to local dwarf / galaxies in dust continuum sed and @xmath13}/l_{\rm fir}$ ] ratio simultaneously . to reveal this
, we need statistically large number of dwarf / irregular galaxy data whose dust continuum seds and @xmath13}/l_{\rm fir}$ ] ratios are available at the same time . on the other hand ,
ly@xmath1 - 1 is in the low redshift lirg region or at the border between the regions of local normal star - forming galaxies and low redshift lirgs in figure [ lciioverlfir_vs_lfir ]
. this could be because it is associated with the special dense environment consisting of a quasar , an smg and other lae and likely a system different from common laes but more similar to lirgs as discussed in [ cii_sfr ] .
as for lbg-1 in figure [ lciioverlfir_vs_lfir ] , it has a lower @xmath14 and/or a higher @xmath13}/l_{\rm fir}$ ] ratio than quasar host galaxies , submm galaxies and @xmath2052 galaxies and is clearly separate from them .
lbg-1 is rather close to the locations of low redshift lirgs and the high @xmath14 edge of local normal galaxies but far apart from the @xmath2027 laes .
it should be noted that as we mentioned earlier , the alma observaton of lbg-1 by @xcite is a factor 37 shallower than those of iok-1 , himiko and hcm 6a .
thus , we can not tell how different the @xmath2027 laes and lbg-1 are in @xmath14 and @xmath13}/l_{\rm fir}$ ] ratio from the current data alone .
however , figure [ lciioverlfir_vs_lfir ] implies that @xmath2027 laes and lbg-1 can not have similar @xmath14 s and similar @xmath13}/l_{\rm fir}$ ] ratios at the same time , because the [ cii ] line is detected from lbg-1 at the @xmath13}$ ] about an order of magnitude higher than the @xmath13}$ ] or @xmath13}$ ] limits of @xmath2027 laes ( see figure [ sfr_vs_lcii ] ) . in any case ,
observations of more lbgs and laes to comparable depth is necessary to fairly compare laes and lbgs in the context of @xmath14 and @xmath13}/l_{\rm fir}$ ] ratio .
after all , the locations of the @xmath1707 laes in the @xmath13}/l_{\rm fir}$]@xmath14 diagram in figure [ lciioverlfir_vs_lfir ] are consistent with what is expected from the locations of these laes in the sfr@xmath13}$ ] diagram in figure [ sfr_vs_lcii ] .
these two diagrams together suggest that the @xmath1987 laes , iok-1 , himiko and hcm 6a are similar to but more likely lower than local normal star - forming galaxies in @xmath13}$ ] , @xmath14 and @xmath13}/l_{\rm fir}$ ] ratio .
this implies that these laes have either lower enriched gas ( or metallicity ) and dust contents or different pdr structures / physical states of ism than local star - forming galaxies .
also , the two diagrams imply that the @xmath202 lae , ly@xmath1 - 1 , could be a system similar to low redshift lirgs .
we have conducted alma 158 @xmath3 m [ cii ] line and fir continuum observation of a @xmath0 lae , iok-1 , a normally star - forming galaxy in the epoch of cosmic reionization , reaching line and continuum sensitivities of @xmath2 240 @xmath3jy beam@xmath4 ( over a channel width of 40 km s@xmath4 ) and @xmath5 21 @xmath3jy beam@xmath4 at a resolution of @xmath217 , comaprable of the size iok-1 in ly@xmath1 .
we did not detect the fir continuum .
we constructed the uv fir sed of iok-1 from its subaru / suprime - cam , _
hst_/wfc3 , _
spitzer_/irac , _ herschel_/spire and our alma data ( @xmath28 upper limit of @xmath218 @xmath3jy ) and compared it with template seds of several types of local galaxies whose fir continua were both corrected and uncorrected for the cmb effects .
we found that iok-1 is similar to local dwarf and irregular galaxies rather than highly obscured / dusty systems . also , modeling the dust continuum sed of iok-1 as the mbb with the dust temperature and the emissivity index typical of dwarf galaxies and scaling it to the cmb - corrected alma fir continuum flux upper limit , we estimated the upper limits on the dust mass , total fir luminosity ( @xmath219 122.5 @xmath3 m ) , total ir luminosity ( @xmath220 @xmath3 m 1 mm ) and dust - obscured sfr of iok-1 to be @xmath7 , @xmath8 , @xmath9 and sfr@xmath221 yr@xmath4 , respectively .
since total sfr of iok-1 is sfr@xmath117 sfr@xmath132 @xmath184 sfr@xmath119 @xmath11 yr@xmath4 , we estimate that @xmath222% of star formation in iok-1 is obscured by dust .
moreover , we found that if we do not correct for the effects of cmb on fir continuum flux measurement , we underestimate the physical quantities related to dust , @xmath93 , @xmath14 , @xmath112 and sfr@xmath123 , by 24% . on the other hand
, we found a @xmath28 weak signal in the [ cii ] image and spectrum at the position of iok-1 and at 238.805 ghz , corresponding to the redshift slightly lower than the one previously estimated from ly@xmath1 emission .
this is consistent with the idea that redshift estimated from ly@xmath1 was overestimated due to the absorption by igm .
since the signal is marginal for securely concluding that it is indeed the [ cii ] line , we have treated it as non - detection and investigated the properties of iok-1 related to [ cii ] line using the @xmath28 upper limit of @xmath223 @xmath3jy ( over a channel width of 40 km s@xmath4 ) .
this corresponds to the [ cii ] luminosity of @xmath13 } < 3.4 \times 10 ^ 7 l_{\odot}$ ] . in the sfr@xmath13}$ ] and the @xmath13}/l_{\rm fir}$]@xmath14 planes , we compared the locations of iok-1 as well as other @xmath1706.6 laes previously observed in the [ cii ] line and fir continuum with those of other objects including local to low redshift normal star - forming galaxies , low redshift lirgs , low to high redshift quasar host galaxies and smgs and a high redshift lbg .
we found that @xmath1987 laes at reionization epoch are systems different from lirgs , quasar hosts , smgs and the lbg but similar to or even lower than local star - forming galaxies in @xmath13}$ ] , @xmath14 and @xmath13}/l_{\rm fir}$ ] ratio .
this implies that these laes have either lower enriched gas ( or metallicity ) and dust or different pdr structures / physical states of ism than those of local normal star - forming galaxies .
the present study is based on only one @xmath224 lae ( as well as a few other @xmath1706.6 laes previously observed in [ cii ] and fir continua ) and results obtained from the analysis of them might not be representative of the [ cii ] line and fir continuum properties of the general population of laes at the epoch of reionization .
fortunately , the number of antennas has been increasing for alma as its construction proceeds towards full capability , substantially improving the sensitivity and resolution .
this will soon allow us to investigate [ cii ] lines and fir continua of the statistically larger number of high redshift star - forming galaxies at a broader redshift range lower and higher than @xmath225 ( epoch of the end of reionization ) in reasonable integration times .
such observations will give us more general insight into the obscured star formation , physical properties , ism state and even dynamics of high redshift star - forming galaxies and will shed light on their relationship with reionization and other galaxy populations .
this paper makes use of the following alma data : ads / jao.alma # 2011.0.00767.s and 2011.0.00115.s .
alma is a partnership of eso ( representing its member states ) , nsf ( usa ) and nins ( japan ) , together with nrc ( canada ) and nsc and asiaa ( taiwan ) , in cooperation with the republic of chile .
the joint alma observatory is operated by eso , aui / nrao and naoj .
we thank rob j. ivison , steve a. eales and matthew w.l .
smith for kindly providing us with the _ herschel_-atlas images around iok-1 .
herschel_-atlas is a project with _ herschel _ , which is an esa space observatory with science instruments provided by european - led principal investigator consortia and with important participation from nasa .
the h - atlas website is http://www.h - atlas.org/. this work was partly supported by the grant - in - aid for the global coe program `` the next generation of physics , spun from universality and emergence '' from the ministry of education , culture , sports , science and technology ( mext ) of japan .
we thank takatoshi shibuya , steven furlanetto and daniel schaerer for their comments on the paper . k.o .
acknowledges the kavli institute fellowship at the kavli institute for cosmology in the university of cambridge supported by the kavli foundation .
k.o . was supported by the alma japan research grant of naoj chile observatory , naoj - alma-0017 .
we thank our referee for useful comments that helped us to improve this paper .
beelen , a. , cox , p. , benford , d. j. , et al .
2006 , , 642 , 694 boone , f. , schaerer , d. , pell , r. , combes , f. , & egami , e. 2007 , , 475 , 513 boselli , a. , gavazzi , g. , lequeux , j. , & pierini , d. 2002 , , 385 , 454 bouwens , r. j. , illingworth , g. d. , oesch , p. a. , et al .
2012 , , 752 , l5 bridge , c. r. , blain , a. , borys , c. j. k. , et al .
2013 , , 769 , 91 cai , z. , fan , x. , jiang , l. , et al .
2011 , , 736 , l28 capak , p. l. , riechers , d. , scoville , n. z. , et al .
2011 , , 470 , 233 carilli , c. l. , riechers , d. , walter , f. , et al .
2013 , , 763 , 120 carniani , s. , marconi , a. , biggs , a. , et al .
2013 , 559 , a29 chabrier , g. 2003 , pasp , 115 , 763 chapman , s. c. , scott , d. , windhorst , r. a. , et al .
2004 , , 606 , 85 colbert , j. w. , malkan , m. a. , clegg , p. e. , et al .
1999 , , 511 , 721 colbert , j. w. , teplitz , h. , francis , p. , et al .
2006 , , 637 , l89 cox , p. , krips , m. , neri , r. , et al .
2011 , , 740 , 63 da cunha , e. , groves , b. , walter , f. , et al .
2013 , , 766 , 13 dale , d. a. , gil de paz , a. , gordon , k. d. , et al .
2007 , , 655 , 863 de breuck , c. , maiolino , r. , caselli , p. , et al .
2011 , , 530 , l8 decarli , r. , walter , f. , carilli , c. , et al .
2014 , , 782 , l17 de looze , i. , baes , m. , bendo , g. j. , cortese , l. , & fritz , j. 2011 , , 416 , 2712 dunlop , j. s. 2013 , in the first galaxies ( astrophysics and space science library , vol .
396 ; berlin : springer - verlag ) , 223 dunne , l. , eales , s. , edmunds , m. , et al .
2000 , , 315 , 115 dwek , e. , staguhn , j. g. , arendt , r. g. , et al .
2011 , , 738 , 36 eales , s. , dunne , l. , clements , d. , et al .
2010 , pasp , 122 , 499 elbaz , d. , daddi , e. , le borgne , d. et al .
2007 , , 468 , 33 finkelstein , s. l. , papovich , c. , ryan , r. e. , jr . ,
2012 , , 758 , 93 finkelstein , s. l. , papovich , c. , dickinson , m. , et al .
2013 , , 502 , 524 gonzlez - lpez , j. , riechers , d. a. , decarli , r. , et al .
2014 , , 784 , 99 hu , e. m. , mcmahon , r. g. , & egami , e. 1996 , , 459 , l53 hu , e. m. , cowie , l. l. , mcmahon , r. g. , et al . 2002a , , 568 , l75 hu , e. m. , cowie , l. l. , mcmahon , r. g. , et al .
2002b , , 576 , l99 hu , e. m. , cowie , l. l. , barger , a. j. , et al .
2010 , , 725 , 394 iono , d. , yun , m. s. , elvis , m. , et al .
2006 , , 645 , l97 ivison , r. j. , swinbank , a. m. , swinyard , b. , et al .
2010 , , 518 , l35 iye , m. , ota , k. , kashikawa , n. , et al .
2006 , , 443 , 186 jiang , l. , egami , e. , mechtley , m. , et al .
2013 , , 772 , 99 kanekar , n. , wagg , j. , ram chary , r. , & carilli , c. 2013 , , 771 , 20 kashikawa , n. , shimasaku , k. , matsuda , y. , et al .
2011 , , 734 , 119 kaufman , m. j. , wolfire , m. g. , hollenbach , d. j. , & luhman , m. l. 1999 , , 527 , 795 kennicutt , r. c. 1998 , , 36 , 189 konno , a. , ouchi , m. , ono , y. , et al .
2014 , , submitted ( arxiv1404.6066 ) kroupa , p. 2001
, , 322 , 231 maiolino , r. , cox , p. , caselli , p. , et al .
2005 , , 440 , l51 maiolino , r. , caselli , p. , nagao , t. , et al .
2009 , , 500 , l1 malhotra , s. , kaufman , m. j. , hollenbach , d. , et al .
2001 , , 561 , 766 matsuda , y. , yamada , t. , hayashino , t. , et al .
2011 , , 410 , l13 mclinden , e. m. , finkelstein , s. l. , rhoads , j. e. et al .
2010 , , 730 , 136 meijerink , r. , & spaans , m. 2005 , , 436 , 397 mentuch cooper , e. , wilson , c. d. , foyle , k. , et al .
2012 , , 755 , 165 momose , r. , ouchi , m. , nakajima , k. , et al .
2014 , , 442 , 110 ohta , k. , matsumoto , t. , maihara , t. , et al .
2000 , , 52 , 557 ohyama , y. , taniguchi , y. , & shioya , y. 2004 , , 128 , 2704 ono , y. , ouchi , m. , shimasaku , k. , et al .
2010 , , 724 , 1524 ono , y. , ouchi , m. , mobasher , b. , et al .
2012 , , 744 , 83 ota , k. , iye , m. , kashikawa , n. , et al .
2008 , , 677 , 12 ota , k. , iye , m. , kashikawa , n. , et al .
2010a , , 722 , 803 ota , k. , ly , c. , malkan , m. a. , et al .
2010b , , 62 , 1167 ouchi , m. , ono , y. , egami , e. , et al
. 2009a , , 696 , 1164 ouchi , m. , mobasher , b. , shimasaku , k. , et al . 2009b , , 706 , 1136 ouchi , m. , shimasaku , k. , furusawa , h. , et al .
2010 , , 723 , 869 ouchi , m. , ellis , r. , ono , y. , et al .
2013 , , 778 , 102 pascale , e. , auld , r. , dariush , a. , et al .
2011 , , 415 , 911 pettini , m. , shapley , a. e. , steidel , c. c. , et al .
2001 , , 554 , 981 pety , j. , beelen , a. , cox , p. , et al .
2004 , , 428 , l21 rangwala , n. , maloney , p. r. , glenn , j. , et al .
2011 , , 743 , 94 rawle , t. d. , egami , e. , bussmann , r. s. , et al .
2014 , , 783 , 59 riechers , d. a. , capak , p. l. , carilli , c. l. , et al .
2010 , , 720 , l131 riechers , d. a. , bradford , c. m. , clements , d. l. , et al .
2013 , , 496 , 329 riechers , d. a. , carilli , c. l. , capak , p. l. , et al .
2014 , , submitted ( arxiv1404.7159 ) robertson , b. e. , furlanetto , s. r. , schneider , e. , et al .
2013 , , 768 , 71 rubin , d. , hony , s. , madden , s. c. , et al .
2009 , , 494 , 647 salim , s. , rich , r. m. , charlot , s. et al .
2007 , , 173 , 267 salpeter , e. e. 1955 , , 121 , 161 shapley , a. e. , steidel , c. c. , pettini , m. , & adelberger , k. l. 2003 , , 588 , 65 shibuya , t. , kashikawa , n. , ota , k. , et al .
2012 , , 752 , 114 shibuya , t. , ouchi , m. , nakajima , k. , et al .
2014 , , 788 , 74 silva , l. , granato , g. l. , bressan , a. , & danese , l. 1998 , , 509 , 103 skibba , r. a. , engelbracht , c. w. , dale , d. , et al .
2011 , , 738 , 89 stacey , g. j. , hailey - dunsheath , s. , ferkinhoff , c. , et al .
2010 , , 724 , 957 steidel , c. c. , adelberger , k. l. , shapley , a. e. , et al .
2000 , , 532 , 170 steidel , c. c. , erb , d. k. , shapley , a. e. , et al .
2010 , , 717 , 289 taniguchi , y. , ajiki , m. , nagao , t. , et al .
2005 , , 57 , 165 vallini , l. , gallerani , s. , ferrara , a. , & baek , s. 2013 , , 433 , 1567 valtchanov , i. , virdee , j. , ivison , r. j. , et al .
2011 , , 415 , 3473 venemans , b. p. , mcmahon , r. g. , walter , f. , et al .
2012 , , 751 , l25 wagg , j. , kanekar , n. , & carilli , c. l. 2009 , 697 , l33 wagg , j. , carilli , c. l. , wilner , d. j. , et al .
2010 , , 519 , l1 wagg , j. , & kanekar , n. 2012a , , 751 , l24 wagg , j. , wiklind , t. , carilli , c. l. , et al . 2012b , , 752 , l30 walter , f. , bertoldi , f. , carilli , c. , et al .
2003 , , 424 , 406 walter , f. , riechers , d. , cox , p. , et al .
2009 , , 457 , 699 walter , f. , decarli , r. , carilli , c. , et al . 2012a , , 486 , 233 walter , f. , decarli , r. , carilli , c. , et al . 2012b , , 752 , 93 wang , r. , wagg , j. , carilli , c. , et al . 2010 , , 739 , l34 wang , r. , wagg , j. , carilli , c. l. , et al . 2013 , , 773 , 44 williams , r. j. , wagg , j. , maiolino , r. , et al . 2014 , , 439 , 2096 willott , c. j. , omont , a. , & bergeron , j. 2013 , , 770 , 13 yang , y. , decarli , r. , dannerbauer , h. , et al . 2012 , , 744 , 178 yang , y. , walter , f. , decarli , r. , bertoldi , f. , weiss , a. , dey , a. , prescott , m. k. m. , & badescu , t. 2014 , , 784 , 171 | we present alma observations of the [ cii ] line and far - infrared ( fir ) continuum of a normally star - forming galaxy in the reionization epoch , the @xmath0 ly@xmath1 emitter ( lae ) iok-1 .
probing to sensitivities of @xmath2 240 @xmath3jy beam@xmath4 ( 40 km s@xmath4 channel ) and @xmath5 21 @xmath3jy beam@xmath4 , we found the galaxy undetected in both [ cii ] and continuum .
comparison of uv fir spectral energy distribution ( sed ) of iok-1 , including our alma limit , with those of several types of local galaxies ( including the effects of the cosmic microwave background , cmb , on the fir continuum ) suggests that iok-1 is similar to local dwarf / irregular galaxies in sed shape rather than highly dusty / obscured galaxies .
moreover , our @xmath6 fir continuum limit , corrected for cmb effects , implies intrinsic dust mass @xmath7 , fir luminosity @xmath8 ( 42.5122.5 @xmath3 m ) , total ir luminosity @xmath9 ( 81000 @xmath3 m ) and dust - obscured star formation rate ( sfr ) @xmath10
@xmath11 yr@xmath4 , if we assume that iok-1 has a dust temperature and emissivity index typical of local dwarf galaxies .
this sfr is 2.4 times lower than one estimated from the uv continuum , suggesting that @xmath12% of the star formation is obscured by dust .
meanwhile , our @xmath6 [ cii ] flux limit translates into [ cii ] luminosity , @xmath13 } < 3.4 \times 10 ^ 7 l_{\odot}$ ] .
locations of iok-1 and previously observed laes on the @xmath13}$ ] vs. sfr and @xmath13}/l_{\rm fir}$ ] vs. @xmath14 diagrams imply that laes in the reionization epoch have significantly lower gas and dust enrichment than agn - powered systems and starbursts at similar / lower redshifts , as well as local star - forming galaxies . |
phase - field models have been introduced by g.caginalp in @xcite ( see also e.g. @xcite for generalizations ) as a relaxation of stefan problems where the development of the sharp interface between phases should be found exactly .
in contrast to that , phase field models operate with phase functions that assume values from -1 ( solid state ) to 1 ( liquid state ) at each spatial point and change sharply but smoothly over the solidification fronts so that the phase interfaces become smoothed . mathematically speaking , such models consist of two coupled parabolic equations describing the temperature and phase fields satisfying initial and boundary conditions .
phase field models are frequently used to describe processes of melting , solidification , evaporation , and condensation , which is directly related to applications such as metal casting , design of cooling systems , cryopreservation of living tissues etc .
( see e.g. @xcite ) .
phase field models are also appropriate for the description of phase changes when modeling co@xmath2 sequestration .
it is assumed that the supercritical carbon dioxide , co@xmath2 pressured to a phase between gas and liquid , is injected into a saline aquifer where it may either dissolve in the brine , react with the dissolved minerals and the surrounding rocks , or become trapped in the pore space of the aquifer . in this
relation , phase field models are useful for the description of transitions between supercritical , ordinary , and dissolved co@xmath2 phases . in this paper , we study the case where a material subjected to a phase change is located inside of a container with heat conductive walls that are not subjected to phase changes . for example , this can be an ampoule filled with a liquid containing living cells ( see ( * ? ? ?
2.2 ) ) and subjected to cooling applied to the outer surface but not immediately to the liquid to be frozen .
the paper is structured as follows .
section [ sec : model ] presents the mathematical model .
the definition of weak solutions and formulation of main results are given in section [ sec : main : res ] .
the existence of weak solutions is shown in section [ sec : galerkin ] . in section [ sec :
energy ] , continuity in time of the phase function and the validity of an energy equality are established .
the uniqueness of weak solutions and their continuous dependence on the initial and boundary data are proved in section [ sec : uniq ] .
first , the phase field model proposed by g.caginalp in @xcite will be recalled , then generalizations considered in this paper will be introduced .
let us outline the model proposed by g.caginalp in @xcite .
assume that a material subjected to phase changes , e.g. liquid to solid and back , occupies a region @xmath3 .
the evolution of the system is described in terms of the temperature @xmath4 and a phase function @xmath5 satisfying the following system of non - linear parabolic equations : @xmath6 the first equation of expresses the balance of heat energy .
notice that this equation is scaled so that the term @xmath7 appears without any multiplier . here , @xmath8 is the scaled heat conductivity coefficient which is assumed to be the same in the solid and liquid states , and @xmath9 is the scaled specific latent heat of the phase change .
the second equation of is derived from statistical physics , namely from the landau - ginzburg theory of phase transitions @xcite . in equilibrium
, @xmath5 minimizes the following free energy functional : @xmath10,\ ] ] where @xmath11 is a length scale , the thickness of the interfacial region between liquid and solid . in non - equilibrium
, @xmath5 does not minimize @xmath12 but satisfies the gradient flow equation @xmath13 , where the symbol @xmath14 denotes the variation of @xmath15 in @xmath5 , which yields the second equation of .
it is clear that the constant @xmath16 characterizes the relaxation time .
the following boundary conditions are usually imposed ( see e.g. @xcite ) : @xmath17 where @xmath18 denotes the outer normal to @xmath19 , and @xmath20 is the scaled overall heat conductivity .
the specific of the case considered in this paper is that the temperature @xmath21 is not directly applied to the boundary of the liquid but to the outer surface of the container ( ampoule ) . to include the container into the model ,
still denote the inner part occupied by the medium subjected to the phase change by @xmath19 , and the solid walls of the container by @xmath22 .
moreover , let @xmath23 be the region occupied by the whole system . in this case , the phase function @xmath5 is defined in @xmath19 , the temperature @xmath4 is defined in @xmath24 , and the boundary function @xmath21 is defined on @xmath25 .
thus , the modification of the model and reads as follows : @xmath26 u_t - k_d \ , \delta u = 0 & \qquad \text { in } d\times(0,t ) , \\
-k_d \ , \partial_{{{\boldsymbol{\nu}}}}u = \lambda(u - g ) & \qquad \text { on } { { \partial u}}\times(0,t ) , \\[1ex ] u({{\boldsymbol{x}}},0)=u^0({{\boldsymbol{x } } } ) & \qquad \text { in } u , \\[1ex ] \phi({{\boldsymbol{x}}},0 ) = \phi^0({{\boldsymbol{x } } } ) & \qquad \text { in } \omega .
\end{aligned}\ ] ] here the indices @xmath22 and @xmath19 denote restrictions to the domains @xmath22 and @xmath19 , respectively , and the symbol @xmath18 is used to indicate outward normals as well to @xmath19 as to @xmath24 , whenever it is not ambiguous . note that the matching conditions imposed on @xmath4 in mean , in particular , the continuity of the temperature and the heat flux across @xmath27 .
the main result of this paper is that the problem admits a unique weak solution @xmath28 , and the mapping @xmath29 ; { l^2(u ) } ) \times \mathcal c([0,t ] ; { h^1(\omega ) } ) \end{split}\ ] ] is continuous provided that @xmath24 and @xmath19 are bounded lipschitz domains in @xmath30 , @xmath31 .
a precise formulation of this result is given in the next section , see definition [ def : wsol ] and theorem [ thm:2reg ] .
problem will be studied in a weak formulation . denote @xmath32 and introduce the following spaces according to @xcite and @xcite : @xmath33 @xmath34 ; { h^1(\omega ) } ) : = \big\ { & \eta\in l^\infty(0,t , { h^1(\omega ) } ) : ~~ t\rightarrow \left<\eta(t ) ; \xi\right>\\ & \mbox{is continuous on~~ } [ 0,t ] \mbox{~~for each~~ } \xi\in ( { h^1(\omega ) } ) ^\prime \big\ } , \end{split } \label{eq : def : cs}\ ] ] where @xmath35 denotes the dual pairing between @xmath36 and @xmath37 . a pair @xmath28 of functions @xmath38 satisfying the initial conditions @xmath39 is a weak solution of problem , if the identities @xmath40 \end{aligned}\ ] ] hold for all test functions @xmath41 and @xmath42 .
[ def : wsol ] [ bemcs ] notice that the initial conditions in definition [ def : wsol ] have a sense because @xmath43 ; { l^2(u ) } ) \ ] ] and @xmath44 ; { h^1(\omega ) } ) , \ ] ] see ( * ? ? ?
25.5 ) for the first assertion and ( * ? ? ?
3 , lemma 8.1 ) for the second one .
the next theorem states the main result of this paper .
[ thm:2reg ] let @xmath24 and @xmath19 be bounded lipschitz domains in @xmath45 with @xmath46 , and @xmath47 be finite . if @xmath48 @xmath49 and @xmath50 , then 1 .
there exists a unique weak solution @xmath28 of problem in the sense of definition [ def : wsol ] .
2 . the phase function @xmath5 has the additional regularity @xmath51 ; { h^1(\omega ) } \right)$ ] .
3 . for all @xmath52 $ ]
, the following energy equation holds : @xmath53 \\ & \quad = { { \int}_\omega}\left [ \frac{1}{8}\phi(s)^4 - \frac{1}{4}\phi(s)^2 + \frac{\xi^2}2|\nabla\phi(s)|^2 \right ] + \int_s^t { { \int}_\omega}\phi_t \left [ 2u- \tau\phi_t \right ] . \end{aligned } \label{eq : energy}\ ] ] 4 .
problem is well - posed in the sense that the mapping @xmath54 ; { h^1(\omega ) } ) \big ) \end{split}\ ] ] is continuous .
the proof of theorem [ thm:2reg ] is given in the following sections .
the existence of weak solutions of problem can be proved in the similar way as theorem 3.1.2 in @xcite , and therefore we only discuss changes in the construction of approximate solutions and a priori estimates . for details related to the passage to the limit , we refer to @xcite .
the next lemma establishes some a priori estimates and the existence of weak solutions to problem .
[ lem : exist ] let @xmath24 and @xmath19 be bounded lipschitz domains in @xmath55 , with @xmath46 , and @xmath47 . if @xmath48 @xmath49 and @xmath50 , then there exists a weak solution @xmath28 to problem in the sense of definition [ def : wsol ] . proof .
let @xmath56 be a basis of @xmath57 which is orthonormal in @xmath58 , and @xmath59 a basis of @xmath36 which is orthonormal in @xmath60 .
consider galerkin approximations of the form @xmath61 where the functions @xmath62 and @xmath63 are to be determined .
let @xmath64@xmath65;l^2({{\partial u}}))$ ] be a sequence such that @xmath66 in @xmath67 as @xmath68 . to determine the functions @xmath69 and @xmath70
, we require that @xmath71 and @xmath72 satisfy the relations @xmath73 \\ & + \int_{\partial u } \lambda \left ( u^m(t)-g^m(t ) \right ) \psi^m , \\ 0 & = { { \int}_\omega}\left [ \tau \ , \phi^m_t(t ) - 2u^m(t ) + \frac{1}{2}\left ( ( \phi^m(t))^3-\phi^m(t)\right ) \right]\eta^m \\ & + { { \int}_\omega}\xi^2\nabla\phi^m(t ) \cdot \nabla\eta^m \end{aligned}\ ] ] for all test functions @xmath74 and @xmath75 .
substituting ansatz and test functions @xmath76 and @xmath77 , @xmath78 , into equations yields the following system of ordinary differential equations for determining @xmath79 and @xmath80 : @xmath81 since the set @xmath82 is orthonormal in @xmath83 and the set @xmath84 is orthonormal in @xmath85 , equations can be rewritten as the following system of odes : @xmath86 \left [ \begin{matrix } \dot{{\boldsymbol{a}}}^m(t)\\ \dot{{\boldsymbol{b}}}^m(t ) \end{matrix } \right ] + \left [ \begin{matrix } { \boldsymbol{a}}^m({{\boldsymbol{a}}}^m(t),\,{{\boldsymbol{b}}}^m(t))\\ { \boldsymbol{b}}^m({{\boldsymbol{a}}}^m(t),\,{{\boldsymbol{b}}}^m(t ) ) \end{matrix } \right ] = { \boldsymbol{0}},\ ] ] where @xmath87 , @xmath88 ; @xmath89 denotes the @xmath90 identity - matrix ; and @xmath91 and @xmath92 denote the terms of which do not comprise the time derivatives of the unknown functions .
initial conditions for @xmath93 and @xmath94 are specified as follows : @xmath95 where @xmath96 and @xmath97 denote the projectors onto the subspaces @xmath98 and @xmath99 in @xmath58 and @xmath36 , respectively .
notice that , for each fixed @xmath100 , the functions @xmath101 and @xmath102 are analytic with respect to all their variables .
thus , the theory of ordinary differential equations provides that , for each @xmath100 , there exist a nonempty time interval @xmath103 $ ] on which the initial value problem and admits a unique solution @xmath104 $ ] . to continue these solutions to any time interval , establish some independent on @xmath100 a priori estimates on @xmath71 and @xmath72 .
first , substitute @xmath105 into the first equation of , integrate it over @xmath106 for some @xmath107 $ ] , and use young s inequality to obtain the estimate @xmath108 for @xmath109 .
next , substitute @xmath110 into the second equation of and proceed as before to obtain @xmath111 \\ & \quad \le \frac \tau 2 { { \int}_\omega}|\phi^0|^2 + \frac{1}{2 } { { \int}_0^t}{{\int}_\omega}|\phi^m|^2 + 2 { { \int}_0^t}{{\int}_u}|u^m|^2 . \end{aligned } \label{eq : estm : phi}\ ] ] now , substitute @xmath112 into the second equation of and proceed as before to obtain @xmath113 \\ & \quad \le { { \int}_\omega}\left [ \frac{1}{8 } \ , |\phi^0|^4 + \frac{\xi^2 } 2 |\nabla\phi^0|^2 \right ] + \frac 1 4 { { \int}_\omega}|\phi^m(t)|^2 + \frac{1}{\epsilon } { { \int}_0^t}{{\int}_u}|u^m|^2 \end{aligned } \label{eq : estm : phit}\ ] ] for @xmath109 .
choosing @xmath114 sufficiently small ; multiplying inequalities , , and by suitable constants ; adding the resulting inequalities ; and using the embedding @xmath115 yield the estimate @xmath116 \\ & \quad + { { \int}_0^t}{{\int}_u}|\nabla u^m|^2 + { { \int}_0^t}{{\int}_{{{\partial u}}}}|u^m|^2 + { { \int}_0^t}{{\int}_\omega}|\phi^m_t(t)|^2 \\ & \quad \le c \left [ 1 + { { \int}_0^t}{{\int}_\omega}|\phi^m|^2 + { { \int}_0^t}{{\int}_u}|u^m|^2 \right ] , \end{aligned } \label{eq : est : gron1}\ ] ] where the constant @xmath117 depends on @xmath16 , @xmath11 , @xmath9 , @xmath20 , @xmath118 , @xmath119 , @xmath120 , and @xmath114 but is independent on @xmath100 . by gronwall s inequality , the left - hand side of is bounded independently on @xmath100 , which along with the embedding @xmath121 , @xmath122 , implies the following assertions : @xmath123 where the bounds are independent on @xmath100 .
this means that the approximate solutions can be continued to any interval @xmath124 $ ] keeping the above mentioned bounds .
now , the proof of the lemma can be completed analogously to that of ( * ? ? ?
in this section , two lemmas are proved .
lemma [ lem : equal : energy ] states a slightly weaker than energy equality , which nevertheless implies the continuity in time of the phase function : @xmath126 ; { h^1(\omega ) } ) $ ] .
lemma [ lem : phi3 t ] completes the proof of the energy equality .
these results will be used in section [ sec : uniq ] to show the claimed uniqueness of weak solutions and their continuous dependence on the initial and boundary data .
the proof of the next lemma is based on the techniques of ( * ? ? ?
* chap 3 , sec 8.4 , lemma 8.3 ) .
[ lem : equal : energy ] let @xmath28 be a weak solution considered in lemma [ lem : exist ] .
then , for all @xmath127 $ ] , the following energy equality holds : @xmath128 = \frac2{\xi^2 } \int_s^t{{\int}_\omega}\phi_t \left [ 2u - \tau\phi_t - \frac{1}{2 } \left(\phi^3 - \phi \right ) \right ] .
\end{aligned } \label{eq : energy : weak}\ ] ] moreover , equality implies that @xmath129;h^1(\omega)\right)$ ] .
assume that holds .
notice that the assertions of provide the integrability of the integrand in the right - hand side of and therefore the convergence of the integral to zero as @xmath130 , which proves the continuity of the function @xmath131 .
this , along with the inclusion @xmath132 ; { l^2(\omega ) } ) $ ] provided by the third and forth assertions of , implies the continuity of the function @xmath133 .
let @xmath134 $ ] be fixed , and @xmath135 as @xmath136 .
denote @xmath137 and compute @xmath138 . utilizing that @xmath139;h^1(\omega))$ ] ( see and remark [ bemcs ] ) implies the convergence @xmath140 , which means that @xmath141 in @xmath142 .
now go on to the proof of equality and introduce some notation .
similar to @xcite , denote the inner product in @xmath60 or the dual pairing in @xmath143 by @xmath144 , and denote the inner product in @xmath145 or the dual pairing in @xmath146 by @xmath147 .
assume that @xmath148 is defined on @xmath149 and possesses the same properties as @xmath148 on @xmath124 $ ] .
this can be achieved trough continuation of @xmath148 by means of reflections . in the following , set @xmath150 and @xmath151 in equation .
define a bilinear form @xmath152 and an operator @xmath153 by @xmath154 then the second equation of is equivalent to the relation @xmath155
\eta .
\label{eq : weak : a}\ ] ] for @xmath156 $ ] and @xmath157 , define the function @xmath158 $ ] by @xmath159 \\ 0 \quad & \mathrm{for } \,\ , t\notin [ 0,t_0 ]
\\ \text{linear on } & [ 0,\delta]\cup[t_0-\delta , t_0]\,\ , \text { and continuous on } \mathbb{r}. \end{cases}\ ] ] let @xmath160 be a sequence of non - negative even regularizing functions with @xmath161 and @xmath162 .
denote @xmath163 .
moreover , as it was already mentioned ( see remark [ bemcs ] ) , @xmath164 ; { h^1(\omega ) } ) .
\label{eq : phi : scalcont}\ ] ] since @xmath165 is a symmetric operator , it holds @xmath166 notice that @xmath167 because of the second equation of and the assertions of .
this implies that @xmath168 since @xmath169 and @xmath170 depend only on time , and @xmath165 is time independent .
therefore , all duality brackets on the right - hand side of equation present the inner product of @xmath171 .
consider the passage to the limit in each term of the right - hand side of as @xmath172 and then as @xmath173 .
the objective is to show the following three relations : @xmath174 , \label{term1lim } \\[2ex ] & \left| \left ( \rho_n \ast a ( q_\delta - q_0 ) \phi \ , ; \ , \rho_n \ast q_\delta ' \phi \right ) \right| \longrightarrow 0 , \label{term2lim } \\[2ex ] & 2 \left ( \rho_n \ast ( a \ , q_0 \ , \phi ) \ , ; \ , \rho_n \ast q_\delta ' \phi \right ) \longrightarrow \left < a\ , \phi(0 ) \ , ; \ , \phi(0 ) \right > - \left < a\ , \phi(t_0 ) \ , ; \ ,
\phi(t_0 ) \right > , \label{term3lim } \end{aligned}\ ] ] as first @xmath172 , and then @xmath173 , which along with equation completes the proof of the lemma .
_ proof of _ : consider the limit as @xmath175 . by properties of convolutions , it holds @xmath176 because these functions have compact supports in @xmath177 .
moreover , @xmath178 in @xmath171 , and therefore , the first term of the right - hand side of the last equation of satisfies @xmath179 as @xmath172 .
relation , the second equation of , and time independence of @xmath165 imply that @xmath180 \end{aligned}\ ] ] as @xmath172 .
moreover , note that @xmath181 in @xmath171 as @xmath173 , because @xmath182 .
therefore , relations and yield .
_ proof of _ : using hlder s inequality yields @xmath183 using again the relation @xmath167 and the properties of the convolution operator , one can show that the first factor on the right - hand side of estimate tends to zero .
the second factor on the right - hand side of can be estimated using young s inequality as follows : @xmath184 the right - hand side is bounded , because @xmath185 , @xmath186 , and @xmath187 $ ] .
thus , the right - hand side of tends to zero as @xmath188 , and relation is proved .
_ proof of _ : notice that @xmath189 ; { l^2(\omega ) } ) $ ] because @xmath190 .
thus , the function @xmath191 is continuous since @xmath192 ; { l^2(\omega ) } ) $ ] .
accounting for the evenness of @xmath169 , the left - hand side of admits the following passage to the limit as @xmath188 : @xmath193 it remains to show that @xmath194 as @xmath173 .
consider the first relation of . by the definition of @xmath170 and @xmath195
, we have @xmath196 and @xmath197 because @xmath195 is even . from the definition of the operator @xmath165 ( see )
, it follows that @xmath198 because the function @xmath199 is continuous , see .
thus , lemma [ lem : equal : energy ] is proved .
@xmath125 the next lemma proves an integral form of the relation @xmath200 , the chain rule .
[ lem : phi3 t ] if @xmath201 ; { h^1(\omega ) } ) $ ] , then @xmath202 = \frac{1}{4 } { { \int}_\omega}\left [ |\phi(t)|^4 - |\phi(s)|^4 \right ] - \frac{1}{2 } { { \int}_\omega}\left [ |\phi(t)|^2 - |\phi(s)|^2 \right ] \label{eq : phi : poly}\ ] ] for all @xmath203 $ ] . [ lem : poly : chain ] proof .
consider the case @xmath150 .
since @xmath204 , it holds @xmath205 , \label{eq : phi2:aim}\ ] ] see e.g. ( * ? ? ?
* th . 1.67 ) . to show the relation @xmath206 , \label{eq : phi4:aim}\ ] ] notice that @xmath207;{l^ { 4 } ( \omega)})$ ] , which follows from ( * ? ? ?
8 , cor 4 ) and the compact embedding @xmath208 for @xmath122 .
let @xmath209 be a sequence of smooth functions such that @xmath210 ; { h^1(\omega ) } ) \quad \text{as } \epsilon \to 0 .
\label{eq : phi4:approx}\ ] ] the chain rule yields @xmath211 .
\label{eq : phi4:eps}\ ] ] the passage to the limit on the both sides of equation , as @xmath212 , has to be done .
the left - hand side is being processed as follows : @xmath213 = { { \int}_0^t}{{\int}_\omega}\left [ \left ( \phi_\epsilon^3 - \phi^3 \right)\ , ( \phi_\epsilon)_t + \phi^3 \left ( ( \phi_\epsilon)_t - \phi_t \right ) \right ] . \end{aligned } \label{eq : phi4:diff}\ ] ] the first summand in the integral on the right - hand side can be estimated using hlder s inequality and the formula @xmath214 with @xmath215 . it holds @xmath216^{1/3 } \cdot \left [ { { \int}_\omega}\left| \zeta^3 \right| \right]^{3/2 } \right\ } { { \mathrm{d}}t}\cdot { { \int}_0^t}{{\int}_\omega}\left| ( \phi_\epsilon)_t \right|^2 \\ & \quad \le c { \left\| \phi_\epsilon - \phi \right\| _ { { l^ { \infty } ( 0,t ; { h^1(\omega ) } ) } } } ^2 \cdot { \left\| \zeta \right\| _ { { l^ { 9/2 } ( 0,t ; { l^ { 3 } ( \omega ) } ) } } } ^{9/2 } \cdot \|(\phi_\epsilon)_t\|^2_{l^2(q_t)}. \end{aligned } \label{eq : phi4:term1}\ ] ] it is not hard to prove tat @xmath217 . really , young s inequality and the embedding @xmath218 yield the estimate @xmath219 ^ 3 \\ & \le { { \int}_\omega}\phi_\epsilon^6 + { { \int}_\omega}\phi^6 + c \sum_{k=1}^5 { { \int}_\omega}\left| \phi_\epsilon^k \ , \phi^{6-k } \right| \\ & \le { { \int}_\omega}\phi_\epsilon^6 + { { \int}_\omega}\phi^6 + c \sum_{k=1}^5 \left [ \frac{k}{6 } { { \int}_\omega}\phi_\epsilon^6 + \frac{6-k}{6 } { { \int}_\omega}\phi^{6 } \right ] \\ & \le c \left [ { \left\| \phi \right\| _ { { l^ { \infty } ( 0,t ; { h^1(\omega ) } ) } } } ^{6 } + { \left\| \phi_\epsilon \right\| _ { { l^ { \infty } ( 0,t ; { h^1(\omega ) } ) } } } ^{6 } \right ] . \end{aligned } \label{eq : zeta : l3}\ ] ]
thus , estimates and yield @xmath220 consider the second summand in the integral on the right - hand side of equation . processing it similarly to the first term yields @xmath221 now , estimates and along with equation yield @xmath222 \longrightarrow 0 \qquad \text{as } \epsilon\to0 .
\label{eq : phi4:diff : lim}\ ] ] using the approximation property , and the embedding @xmath218 yields the estimate @xmath223 = { { \int}_\omega}\left [ \left ( \phi(t ) - \phi_\epsilon(t ) \right )
\zeta(t ) \right ] \\ & \quad \le { \left\| \phi(t ) - \phi_\epsilon(t ) \right\|_{l^ { 2 } ( \omega ) } } \cdot { \left\| \zeta(t ) \right\|_{l^ { 2 } ( \omega ) } } \\ & \quad \longrightarrow 0 \end{aligned } \label{eq : phi4:diff4:lim}\ ] ] for all @xmath224 $ ] as @xmath212 .
here we have used the abbreviation @xmath225 ; { l^2(\omega ) } ) $ ] .
thus , equation follows from equation and relations and , which proves lemma [ lem : phi3 t ] .
@xmath125 finally , the energy equality follows from lemmas [ lem : equal : energy ] and [ lem : poly : chain ] .
the next lemma completes the proof of theorem [ thm:2reg ] .
let @xmath24 and @xmath19 be bounded lipschitz domains in @xmath45 with @xmath46 , and @xmath47 .
let @xmath226 , @xmath227 , and @xmath228 be given initial and boundary data of problem ; @xmath229 weak solutions of problem in the sense of definition [ def : wsol ] . if @xmath230 and @xmath231 , then @xmath232 ; { l^2(u ) } ) } \\ & { \left\| \bar u \right\| _ { { l^ { 2 } ( 0,t ; { h^1(u ) } ) } } } \\ & \|\bar \phi\|_{\mathcal c([0,t ] ; { h^1(\omega ) } ) } \\ & \|\bar \phi\|_{h^1(0,t ; { l^2(\omega ) } ) } \end{aligned } \right\ } \le f\big ( \|\bar u^0\|_{l^2(u ) } , \ , \|\bar\phi^0\| _ { { h^1(\omega ) } } , \ , \|\bar g\|_{l^2({{\partial u}}\times(0,t))}\big ) , \label{eq : cont}\ ] ] where @xmath233 is a continuous function with @xmath234 .
[ lem : uniq ] proof . denote @xmath235 and bear in mind that @xmath236 is non - negative . obviously , @xmath237 and @xmath238 satisfy the equations @xmath239
\end{aligned}\ ] ] for all test functions @xmath41 and @xmath240 .
in particular , test functions @xmath241 , @xmath242 , and @xmath243 will be considered to obtain three estimates . here
, @xmath244 denotes the characteristic function of the interval @xmath106 .
notice that the proofs of lemmas [ lem : equal : energy ] and [ lem : poly : chain ] show that @xmath243 is an admissible test function . substituting @xmath245 into the first equation of and using young s inequality yield the estimate @xmath246 \end{aligned}\ ] ] for any @xmath109 . substituting @xmath242 into the second equation of and applying young s inequality yield @xmath247 \\ & \quad \le \frac \tau 2 { { \int}_\omega}|\bar\phi^0|^2 + { { \int}_0^t}{{\int}_\omega}\left [ |\bar u|^2 + \frac 3 2 |\bar\phi|^2 \right ] . \end{aligned}\ ] ] substituting @xmath243 into the second equation of and applying young s inequality yield @xmath248 + \int_0^{t}{{\int}_\omega}\frac{\tau}{4 } \ , |\bar\phi_t|^2 \\ & \quad \le { { \int}_\omega}\left [ \frac{1}{4}|\bar\phi(t)|^2 + \frac{\xi^2}2|\nabla\bar\phi^0|^2 \right ] + { { \int}_0^t}{{\int}_\omega}\left [ \frac 2 \tau \ ,
\bar u^2 + \frac 1 { 4\tau } \ , \zeta^2\ , \bar\phi^2 \right ] .
\end{aligned}\ ] ] using hlder s inequality and the embedding @xmath218 yield @xmath249 \le { { \int}_0^t}\left\ { \left [ { { \int}_\omega}|\zeta|^3 \right]^{2/3 } \cdot \left [ { { \int}_\omega}|\bar\phi|^6 \right]^{3 } \right\ } \\ & \quad \le { \left\| \zeta \right\| _ { { l^ { \infty } ( 0,t ; { l^ { 3 } ( \omega ) } ) } } } ^2 { { \int}_0^t } { \left\| \bar\phi(t ) \right\|_{l^ { 6 } ( \omega ) } } ^2 \\ & \quad \le c { \left\| \zeta \right\| _ { { l^ { \infty } ( 0,t ; { l^ { 3 } ( \omega ) } ) } } } ^2 \cdot { { \int}_0^t}{{\int}_\omega}\left [ |\bar\phi(t)|^2 + |\nabla\bar\phi(t)|^2 \right ] . \end{aligned } \label{eq : zeta : bar}\ ] ] now , choose @xmath250 in and combine estimates , , , and to obtain @xmath251 \\ & \quad + c \left [ { { \int}_0^t}{{\int}_u}|\bar u|^2 + c_\zeta^2
{ { \int}_0^t}{{\int}_\omega}\left [ |\bar\phi|^2 + |\nabla\bar\phi|^2 \right ] \right ] , \end{aligned } \label{eq : bar : est1}\ ] ] where @xmath252 , and the constant @xmath117 depends on @xmath253 and @xmath16 only . to eliminate the first term on the right - hand side of , multiply inequality by a constant greater than @xmath254 and
add the resulting inequality to ( remember that @xmath236 is non - negative ) .
this yields @xmath255 \\ & \quad + { { \int}_0^t}{{\int}_u}|\nabla\bar u|^2 + { { \int}_0^t}{{\int}_\omega}|\bar\phi_t|^2 + { { \int}_0^t}{{\int}_{{{\partial u}}}}|\bar u|^2 \\ & \quad \le c \left\ { c_0 + ( 1+c_\zeta^2 ) { { \int}_0^t}\left ( { { \int}_u}|\bar u|^2 + { { \int}_\omega}\left [ |\bar\phi|^2 + |\nabla\bar\phi|^2 \right ] \right ) \right\ } , \end{aligned } \label{eq : bar : est2}\ ] ] where @xmath117 is sufficiently large , and @xmath256 + { { \int}_0^t}{{\int}_{{{\partial u}}}}|\bar g|^2 .
\label{eq : def : c0}\ ] ] applying gronwall s inequality to yields @xmath257 \\ & \quad + { { \int}_0^t}{{\int}_u}|\nabla\bar u|^2 + { { \int}_0^t}{{\int}_\omega}|\bar\phi_t|^2 + { { \int}_0^t}{{\int}_{{{\partial u}}}}|\bar u|^2 \\ & \quad \le c_0 \ , c \left [ 1 + t \ , c ( 1+c_\zeta^2 ) \exp \left ( t \ , c ( 1+c_\zeta^2 ) \right ) \right ] . \end{aligned } \label{eq : bar : est3}\ ] ] due to the a priori estimates derived in section [ sec : galerkin ] ( see the assertions of ) , the constant @xmath258 is bounded in terms of the problem data only .
thus , estimate proves the continuous dependence of weak solutions on the problem data .
in particular , inequality and the definition of @xmath259 ( see ) imply uniqueness of weak solutions.@xmath125
the following simulation deals with a plastic ampoule used for freezing small tissue samples ( see @xcite ) .
the diameter of the ampoule is equal to 1 cm , the height equals 5 cm , and the wall thickness equals 0.1 cm . the ampoule is filled with water , and is being cooled with the rate of 1@xmath260c / s applied to the outer surface of the ampoule .
the mass and thermal characteristics of plastic and water are easily available in reference books .
the length scale @xmath11 and the relaxation time @xmath16 are equal to @xmath261 and @xmath262 , respectively .
the first picture of figure [ fig1 ] shows an axial cut of the whole ampoule ( region @xmath24 ) .
the others show the water part ( region @xmath19 ) only .
the unfrozen part is shown in light , whereas the frozen one becomes dark .
( 9,5 ) ( 0,0 ) ( 0.2,0 ) in the light region , and @xmath263 in the dark one .
there is a small transition zone between the unfrozen and frozen parts.,title="fig:",height=188 ] ( 3.0,0 ) in the light region , and @xmath263 in the dark one .
there is a small transition zone between the unfrozen and frozen parts.,title="fig:",height=188 ] ( 5.3,0 ) in the light region , and @xmath263 in the dark one .
there is a small transition zone between the unfrozen and frozen parts.,title="fig:",height=188 ] ( 7.9,0 ) in the light region , and @xmath263 in the dark one .
there is a small transition zone between the unfrozen and frozen parts.,title="fig:",height=188 ] ( 10.5,0 ) in the light region , and @xmath263 in the dark one .
there is a small transition zone between the unfrozen and frozen parts.,title="fig:",height=188 ] ( 0,3)@xmath264 ( 2.8,3)@xmath265 ( 5.2,3)@xmath266 ( 7.6,3)@xmath267 ( 10.3,3)@xmath268
we have considered a phase field model describing phase changes of a medium located in a container with heat conductive walls that are free of phase changes .
it is shown that the temperature and the phase variables are continuous functions with values in @xmath83 and @xmath142 , respectively , provided that the initial data are from @xmath83 and @xmath142 , respectively .
moreover , solutions depend continuously on the initial and boundary data of the problem . +
* acknowledgements . *
the work was motivated by the project spp 1253 of the german research society ( dfg ) , and was supported by award no .
ksa - c0069/uk - c0020 , made by king abdullah university of science and technology ( kaust ) .
a. novotny and i. straskraba .
_ introduction to the mathematical theory of compressible flow _ ,
volume 27 of _ oxford lecture series in mathematics and its applications_. oxford university press , august 2004 . | a phase field model proposed by g.caginalp for the description of phase changes in materials is under consideration .
it is assumed that the medium is located in a container with heat conductive walls that are not subjected to phase changes .
therefore , the temperature variable is defined both in the medium and wall regions , whereas the phase variable is only considered in the medium part . the case of lipschitz domains in two and three dimensions
is studied .
we show that the temperature and phase variables are continuous in time functions with values in @xmath0 and @xmath1 , respectively , provided that the initial values of them are from @xmath0 and @xmath1 , respectively .
moreover , continuous dependence of solutions on the initial data and boundary conditions is proved . |
studying the properties of the gaseous environments of radio galaxies and quasars on different scales are important for understanding the interactions of the radio jets with the external environment and the evolution of these sources .
such studies could also provide useful insights towards understanding the triggering of radio activity , and examining consistency of these properties with the unified schemes for active galactic nuclei ( agn ) ( pihlstrm , vermeulen & conway 2003 ; gupta & saikia 2006b ) .
a useful way of probing the cold neutral component of this gas is via 21-cm hi absorption towards radio sources , which range in size from the sub - galactic sized compact steep - spectrum ( css ) and gigahertz peaked - spectrum ( gps ) sources , to the large radio galaxies and quasars which are up to a few mpc in size .
the css and gps objects ( o dea 1998 ) have been inferred to be young ( @xmath1110@xmath12 yr ) , while the larger sources could be older than @xmath810@xmath13 yr ( jamrozy et al .
2008 ; konar et al .
2008 ) .
several hi absorption line studies have shown that css and gps objects tend to exhibit absorption lines more frequently than larger sources , with the lines being both blue- and red - shifted relative to the systemic velocity and exhibiting complex line profiles . the hi column densities also appear to be anticorrelated with the source sizes ( pihlstrm , conway & vermeulen 2003 ; gupta et al .
2006 ) , and are broadly consistent with the unified scheme for radio galaxies and quasars .
the relationship between the hi column density and core prominence , which is being used as a statistical indicator of source orientation , appears to be consistent with the hi gas being distributed in a circumnuclear disk on a scale smaller than the size of the compact radio sources ( gupta & saikia 2006b ) .
evidence of absorption arising from a circumnuclear disk - like structure has also been inferred from higher - resolution , very long baseline interferometric ( vlbi)-scale spectroscopic observations in several sources such as the css object j0119 + 3210 ( 4c+31.04 ) and the cores of a few larger sources , cyga , ngc4261 and hydraa ( e.g. conway & blanco 1995 ; taylor 1996 ; conway 1999 ; van langevelde et al . 2000 ) .
a comparison of the hi absorption properties towards the cores of larger sources with those of the css and gps objects might provide insights towards understanding the evolution of the gaseous properties as the source ages .
such a study indicates that the detection of hi in absorption towards the cores of larger sources is significantly smaller than for css and gps objects suggesting an evolution in the gaseous content of the host galaxies with source age ( chandola et al .
, in preparation )
. there have also been suggestions that the torus / disk may be different in fri and frii sources , which is likely to be reflected in the hi absorption properties of these two classes of sources ( e.g. morganti et al .
given the small number of detections at present , the difference in detection rate between fri and frii sources does not appear to be significant ( chandola et al .
, in preparation ) .
an important aspect in our understanding of agn is the episodic nature of their nuclear activity , and the physical processes that might be governing it ( saikia & jamrozy 2009 for a review )
. there appears to be a trend for a high rate of detection of hi absorption in sources with evidence of rejuvenated activity .
these include the giant radio galaxy 3c236 which also exhibits evidence of star formation ( conway & schilizzi 2000 ) , the giant radio galaxy j1247 + 6723 with a gps core ( saikia , gupta & konar 2007 ) , the misaligned ddrg 3c293 ( beswick et al .
2004 ; emonts et al . 2005 ) , the well - studied southern radio galaxy centaurus a ( sarma , troland & rupen 2002 ; morganti et al . 2008 ) , 4c 29.30 ( chandola , saikia & gupta 2010 ) and cta 21 ( salter et al . 2010 ) .
the well - known frii radio galaxy cygnus a , which has been shown to have two cycles of radio activity from radio and x - ray observations ( steenbrugge , blundell & duffy 2008 ; steenbrugge , heywood & blundell 2010 ) , also exhibits nuclear hi absorption ( conway 1999 ) . in this paper
we present hi observations with the giant metrewave radio telescope ( gmrt ) to localise and put constraints on the size of the hi absorber in the frii radio galaxy 3c321 , which has been reported earlier from arecibo observations ( mirabel 1990 ) .
we also present radio continuum observations with the gmrt and the very large array ( vla ) of the diffuse plume of emission to explore whether this feature might be relic emission from an earlier cycle of activity .
the radio galaxy 3c321 is an frii radio galaxy at a redshift of 0.0961 ( luminosity distance = 436 mpc ) so that 1 arcsec corresponds to 1.759 kpc in a universe with h@xmath14=71 km s@xmath0 mpc@xmath0 , @xmath15=0.27 and @xmath16=0.73 .
our observations show that it has two prominent hot - spots at the outer edges with an overall angular size of @xmath8 286 arcsec , which corresponds to a projected linear size of @xmath8 503 kpc , similar to that found in the literature ( e.g. baum et al .
1988 ; lal , hardcastle & kraft 2008 ) . on smaller scales ,
a vla image with an angular resolution of @xmath81.4 arcsec by baum et al .
( 1988 ) , shows a compact core , a knot towards the north - west separated from the core by @xmath83.5 arcsec ( 6 kpc ) , and a somewhat diffuse but bright and prominent jet extending upto @xmath838 kpc .
there is also evidence of a weaker counter - jet towards the south - east .
a lower - resolution radio image has shown evidence of collimated emission extending all the way to the north - western hot - spot , and a diffuse plume of emission starting beyond the bright emission in the jet and extending for @xmath8140 kpc at an angle of @xmath830@xmath9 to the source axis ( evans et al .
2008 ) .
optical observations show that the galaxy associated with the radio source has a prominent kpc - scale dust lane , and a fainter companion towards the north - west ( hurt et al .
1999 ; martel et al . 1999
; roche & eales 2000 ; de koff et al . 2000 ) .
the two galaxies are embedded in a common stellar envelope , and the velocity difference between the two galaxies is within @xmath8200 km s@xmath0 .
optical spectroscopic and x - ray observations suggest that the companion galaxy also hosts an agn ( filippenko 1987 ; robinson et al .
2000 ; evans et al . 2008 ) .
there are different regions of x - ray emission , with emission from both the hot - spots and regions offset from the radio peaks ( hardcastle et al .
2004 ; evans et al .
hubble space telescope ( _ hst _ ) stis nuv and [ o ] images , along with the _ chandra _ and _ spitzer _ irac 4.5 @xmath17 m images show the complex distribution of gas in the circumnuclear region ( e.g. evans et al . 2008 ) .
the results presented here are based on hi observations made with the gmrt , as well as on archival gmrt and vla data .
the observing log for both the gmrt and vla observations is listed in table 1 .
.the observation log .
column 1 : the name of the telescope , and the array configuration for the vla observations ; column 2 : dates of the observations ; columns 3 and 4 : the frequency and bandwidth used in making the images . [ cols="^,^,^,^,^,^ " , ] @xmath18 flux density has been estimated over same area as the vla images , and the total flux density of the plume is likely to be higher . the spectrum of the plume between 1511 and 4860 mhz , estimated from similar resolution vla images , is steep with a spectral index of @xmath80.8 . the flux density of the plume at 614 mhz has been estimated over same area as at the higher frequencies . the plume is seen over a smaller extent at 614 mhz compared with the vla images , and our estimate of the flux density at this frequency is @xmath8100 mjy .
lal , hardcastle & kraft ( 2008 ) estimate the flux density of the plume which they term the ` north wing ' to be 32.1 and 17.2 mjy from their high- and low - resolution images at 610 mhz , and a spectral index of 0.30 between 1500 and 610 mhz .
our flux density estimate is clearly significantly higher than those of lal et al .
we imaged the archival gmrt data at the lower frequencies to determine the spectrum of the plume , but the images were not of satisfactory quality to give reliable estimates of the flux densities of the plume ( see for example the 240-mhz image of lal et al .
the plume is also barely seen in the tifr gmrt sky survey ( tgss ; sirothia et al .
, in preparation ) image at 148 mhz ; clearly better quality data are required to determine the low - frequency spectrum of the plume reliably .
this would help determine any break in the spectrum and explore the possibility of whether the plume might be relic emission due to an earlier cycle of activity .
although the present suggestion is that the plume is caused by the deflection of the jet ( e.g. evans et al .
2008 ) , the detailed spectrum of the plume would help examine the alternate possibility of this being relic emission .
if the plume of emission is indeed due to an earlier cycle , the detection of hi in absorption would be consistent with the trend for a high incidence of absorption in sources with evidence of recurrent activity ( cf .
saikia & jamrozy 2009 for a review ; salter et al .
2010 and references therein ) .
the gmrt hi observations show that the absorption occurs towards the core component , with also a suggestion of absorption towards the knot in the jet 6 kpc north - west of the nucleus .
this would indicate the size of the absorber to be @xmath19 6 kpc if it is a single absorber .
the absorption profile towards the core has multiple components with a total column density of @xmath1(hi)=9.23@xmath210@xmath3(@xmath4/100)(@xmath7/1.0 ) @xmath6 .
the column density towards the knot is @xmath1(hi)=7.54@xmath210@xmath3(@xmath4/100)(@xmath7/1.0 ) @xmath6 .
no absorption is seen towards the hot - spots .
radio continuum observations of lobes with similar resolution show that their spectra are straight up to 5 ghz , suggesting spectral ages of @xmath1026 myr for the lobes .
the detection of x - rays from the hot - spots suggests that these are being continuously fed by the jets , consistent with their spectral ages .
the full extent of the plume is not seen in our 614-mhz gmrt and also barely seen in the tgss image at 148 mhz
. however , better quality data are required to determine the spectrum of the plume reliably over a large frequency range to try and examine a possible break in the spectrum and explore whether the plume might be relic emission from an earlier cycle of activity .
we thank george privon for useful discussions and sharing his results on 3c321 with us , dave green , the reviewer for many useful comments which helped improve the presentation of the paper , and the staff of gmrt for their help with the observations .
the gmrt is a national facility operated by the national centre for radio astrophysics of the tata institute of fundamental research .
we thank the nrao for use of the archival data .
the national radio astronomy observatory is a facility of the national science foundation operated under co - operative agreement by associated universities inc .
this research has made use of the nasa / ipac extragalactic database ( ned ) which is operated by the jet propulsion laboratory , caltech , under contract with the national aeronautics and space administration .
baars j.w.m . , genzel r. , pauliny - toth i.i.k . , witzel a. , 1977 , a&a , 61 , 99 baum s.a . , heckman t.m . ,
bridle a. , van breugel w.j.m . , miley g. k. , 1988 , apjs , 68 , 643 beswick r.j . , peck a.b .
, taylor g.b . , giovannini g. , 2004 , mnras , 352 , 49 chandola y. , saikia d.j . , gupta n. , 2010 , mnras , 403 , 269 conway j.e . , 1999 ,
aspc , 156 , 259 conway j.e . ,
blanco p.r .
, 1995 , apj , 449 , l131 conway j.e .
, schilizzi r.t . , 2000 , proceedings of the 5th european vlbi network symposium , p. 123 de koff s. , et al .
2000 , apjs , 129 , 33 emonts b.h.c .
, morganti r. , tadhunter c.n .
, oosterloo t.a . ,
holt j. , van der hulst j.m .
, 2005 , mnras , 362 , 931 evans d.a . , et al . 2008 , apj , 675 , 1057 filippenko a.v . , 1987 , iaus , 124 , 761 giovannini g. , feretti l. , gregorini l. , parma p. , 1988 ,
a&a , 199 , 73 gupta n. , saikia d.j . , 2006a , mnras , 370 , l80 gupta n. , saikia d.j . , 2006b , mnras , 370 , 738 gupta n. , salter c.j .
, saikia d.j . , ghosh t. , jeyakumar s. , 2006 , mnras , 373 , 972 hurt t. , antonucci r. , cohen r. , kinney a. , krolik j. , 1999 , apj , 514 , 579 hardcastle m.j .
, harris d.e .
, worrall d.m . , birkinshaw m. , 2004 , apj , 612 , 729 jamrozy m. , konar c. , machalski j. , saikia d.j . , 2008 , mnras , 385 , 1286 konar c. , jamrozy m. , saikia d.j . , machalski j. , 2008 , mnras , 383 , 525 lal d.v . ,
hardcastle m.j .
, kraft r.p . , 2008 , mnras , 390 , 1105 martel a.r . , et al . 1999 , apjs , 122 , 81 miley g. , 1980 , ara&a , 18 , 165 mirabel i. f. , 1990 , apj , 352 , l37 morganti r. , oosterloo t.a .
, tadhunter c.n .
, van moorsel g. , killeen n. , wills k.a .
, 2001 , mnras , 323 , 331 morganti r. , tadhunter c.n .
, oosterloo t.a . , 2005 , a&a , 444 , l9 morganti r. , oosterloo t. , struve c. , saripalli l. , 2008 , a&a , 485 , l5 odea c.p . , 1998 , pasp , 110 , 493 pihlstrm y. , conway j. , vermeulen r. , 2003 , pasa , 20 , 62 robinson t.g .
, tadhunter c.n .
, axon d.j . ,
robinson a. , 2000 , mnras , 317 , 922 roche n. , eales s.a . , 2000 , mnras , 317 , 120 saikia d.j . ,
jamrozy m. , 2009 , basi , 37 , 63 saikia d.j . , gupta n. , konar c. , 2007 , mnras , 375 , l31 sarma a.p .
, troland t.h .
, rupen m.p . , 2002 ,
apj , 564 , 696 salter c.j .
, saikia d.j . , minchin r. , ghosh t. , chandola y. , 2010 , apj , 715 , l117 steenbrugge k.c . , blundell k.m . , duffy p. , 2008 ,
mnras , 388 , 1465 steenbrugge k.c . , heywood i. , blundell k.m . , 2010 , mnras , 401 , 67 tadhunter c. , wills k. , morganti r. , oosterloo t. , dickson r. , 2001 , mnras , 327 , 227 taylor g.b .
, 1996 , apj , 470 , 394 vermeulen r. c. , et al . 2003 , a&a , 404 , 861 vermeulen r.c . , labiano a. , barthel p.d . ,
baum s.a . , de vries w.h . , odea c.p . , 2006 ,
a&a , 447 , 489 van langevelde h.j .
, pihlstrm y.m . , conway j.e . , jaffe w. , schilizzi r.t . , 2000 ,
a&a , 354 , l45 | we report the results of giant metrewave radio telescope ( gmrt ) observations of hi absorption towards the frii radio galaxy 3c321 ( j1531 + 2404 ) , which is associated with an active galaxy interacting with a companion .
the absorption profile towards the radio core is well resolved and consists three components , of which the two prominent ones are red - shifted by 186 and 235 km s@xmath0 relative to the optical systemic velocity . the neutral hydrogen column density towards the core is estimated to be @xmath1(hi)=9.23@xmath210@xmath3(@xmath4/100)(@xmath5/1.0 ) @xmath6 , where @xmath4 and @xmath7 are the spin temperature and covering factor of the background source respectively .
we also present radio continuum observations of the source with both the gmrt and the very large array ( vla ) in order to understand the properties of a plume of emission at an angle of @xmath830@xmath9 to the source axis .
this feature appears to have a steep high - frequency spectrum .
the current hotspots and jet are active and seen in x - ray emission .
the spectral ages of the lobes are @xmath1026 myr .
we discuss the possibility that the plume could be relic emission due to an earlier cycle of activity .
[ firstpage ] galaxies : active galaxies : nuclei galaxies : individual : 3c321 radio continuum : galaxies radio lines : galaxies |
in recent years , the process of the false vacuum decay with the effect of gravity has been studied extensively in the context of the one - bubble open inflation scenario , in which an open universe is created in a nucleated bubble formed through false vacuum decay@xcite . in this context
, we are interested in the spectrum of quantum fluctuations after the bubble nucleation because it determines the spectrum of primordial fluctuations of the universe . by comparing the predicted spectrum with the observed one
, we can test whether a certain model is viable or not .
the fluctuation can be decomposed by using the spatial harmonics in an open universe .
we denote the eigen value of the spatial harmonics by @xmath0 .
the spatial harmonics with positive @xmath0 are square integrable on a time constant hypersurface in an open universe , and we have a continuous spectrum for @xmath1 . while , the spatial harmonics are no longer square integrable when the eigen value @xmath0 is negative . however , since any time constant hypersurface in an open universe is not a cauchy surface , this divergence does not directly prohibit the existence of such a mode . by considering the normalization of perturbation modes on an appropriate cauchy surface , we find that the spectrum for @xmath2 can exist as a discrete one , which we call supercurvature mode@xcite .
further , we classified the modes with negative @xmath0 into two classes depending on their eigen values .
one is supercritical modes with @xmath3 and the other is subcritical modes with @xmath4 . if and only if there exist supercritical modes , the two - point correlation function of the tunneling field unboundedly increases for large spatial separation in the open universe created in the nucleated bubble .
the diverging correlation for large spatial separation does not seem to be allowed intuitively .
furthermore , it is known that we meet a trouble in constructing homogeneous fluctuations for supercritical modes except for some restricted cases@xcite .
hence , we proposed the `` no - supercritical supercurvature mode conjecture '' , which is the conjecture claiming that the existence of supercritical supercurvature modes should be prohibited in realistic models of creation of an open universe@xcite .
on the other hand , there is an unsolved problem in the euclidean path integral approach to describe the phenomena of the true vacuum bubble nucleation with the effect of gravity through quantum tunneling@xcite .
that is the so - called negative mode problem@xcite . in the lowest wkb approximation ,
the quantum tunneling is described by a bounce solution@xcite .
the decay rate per unit volume and per unit time interval , @xmath5 , is given by the formula @xmath6 where @xmath7 is
the classical euclidean action for the bounce solution and @xmath8 is that for the trivial solution that stays in the false vacuum . in the path integral
approach , the prefactor @xmath9 is evaluated by taking the imaginary part of the gaussian integral over fluctuations around the background bounce solution . for an ordinary system without gravity , there is one perturbation mode in which direction the action does not decrease .
it is called negative mode .
the existence of an unique negative mode has already proved for quantum fields in flat spacetime@xcite . in evaluating this gaussian integral ,
the path of integration is deformed on the complex plane to make the integral well - defined . as a result , one imaginary factor , @xmath10 , appears in @xmath11 . in the euclidean path integral
approach to the tunneling , this imaginary unit plays a crucial role when we interpret @xmath5 as the decay rate .
however , in the case where gravity is taken into account , the situation is different . in this case , the tunneling is described by the @xmath12-symmetric coleman and de luccia ( cd ) bounce solution@xcite . in reducing the action for fluctuations around the @xmath12-symmetric cd bounce solution , we used the standard gauge fixing method to deal with gauge degrees of freedom@xcite .
then , we obtained the reduced action that retains only the physical degrees of freedom@xcite . reflecting
the fact that the action with gravitational degrees of freedom is unbounded below , the reduced action for the fluctuations that conserve the @xmath12-symmetry has an unusual overall signature . to treat this system
, we proposed to use the prescription similar to the conformal rotation@xcite .
then , from path integral measure , there arise one imaginary unit @xmath10 . hence , in order to obtain a correct number of @xmath10 , i.e. , one , it seems that there should exist no - negative mode .
therefore , we proposed the `` no - negative mode conjecture '' .
we have shown in our previous paper@xcite that the existence of supercritical supercurvature modes is equivalent to the existence of negative modes .
that is , if we obtain a proof of the `` no - negative mode conjecture '' , the `` no - supercritical supercurvature mode conjecture '' is also proved .
henceforce , the issue if the `` no - negative mode conjecture '' is true or not is now of increasing importance . in the same paper@xcite
, we have shown that this conjecture holds in a certain restricted situation . in the present paper
, we give a proof of this conjecture for rather general models which consist of one scalar field .
here , we briefly describe the statement proved in the present paper . here
, we should note that not all cd bounce solutions contribute to the tunneling process .
the relevant contribution to the decay rate comes only from the bounce solution that realizes the minimum value of action among all non - trivial solutions .
thus , the statement that we should prove is the following : `` there is no - negative mode for the cd bounce solution that realizes the minimum value of action among all non - trivial solutions . ''
however , making a list of all the cd bounce solutions , we find that there are many solutions for which the tunneling field does not change monotonically .
these solutions have more than two concentric domain walls .
we expect that the simplest bounce solution with one - domain wall will dominate the tunneling process .
hence , one may want to strengthen the conjecture as follows ; for the bounce solution that gives the minimum value of action , the tunneling field @xmath13 changes monotonically , and this solution have no - negative mode .
we refer to this version of conjecture as the `` no - negative mode theorem '' .
we used the word
theorem instead of ` conjecture ' because this statement is proved in the this paper .
this paper is organized as follows . in section
ii , we explain the `` no - negative mode theorem '' in more detail . in section iii , we present a prescription to search for all cd bounce solutions . in section iv , we give a method to count the number of negative modes for a given cd bounce solution . in section
v , we give a method to count the multiplicity of domain walls contained in a given soltion . in section
vi , we give a method to compare the values of action between various cd bounce solutions for a given potential model . in section vii , combining the results obtained in sec .
iv , v and vi , we prove the `` no - negative mode theorem '' .
we summarize the outline of the proof and discuss the implication of this theorem in section viii .
in this section , we explain the statement , which we call the `` no - negative mode theorem '' , in more detail to clarify what we are going to prove .
we consider the system composed of a scalar field , @xmath14 , with the einstein gravity .
the euclidean action is given by @xmath15,\ ] ] where @xmath16 and @xmath17 is the ricci curvature .
the potential of the scalar field is assumed to have the form as shown in fig.[fig1 ] , and initially the field is trapped in the false vacuum .
we set @xmath18 at the top of the potential barrier and we denote the bottoms in the false and true vacua by @xmath19 and @xmath20 , respectively .
we assume that @xmath21 vanishes only at @xmath20 , @xmath22 and @xmath19 .
the bounce solution is an euclidean solution which connects the configurations before and after tunneling . in the present case , before tunneling the geometry is given by the de sitter space and the scalar field is in the false vacuum . after tunneling
, there appears a true vacuum bubble in the false vacuum sea . under the assumption of the @xmath12-symmetry , @xmath23
the corresponding bounce solution is found by coleman and de luccia@xcite . with this assumption
, the euclidean action reduces to @xmath24 , \label{newaction}\ ] ] [ fig1 ] where @xmath25 represents the differentiation with respect to @xmath26 and @xmath27 . here
@xmath28 is the value of @xmath26 at both boundaries .
we assume that @xmath29 and @xmath30 . for definiteness , we assume that @xmath13 is in the true ( false ) vacuum side at @xmath31 . due to this restriction
, we do not take account of solutions in which @xmath13 is in the same vacuum side at both boundaries .
the euclidean equations of motion are obtained by taking the variation of @xmath32 . after taking variations
, we adopt the synchronous gauge by setting @xmath33 , for simplicity . the differentiation with respect to @xmath34
gives a constraint equation , @xmath35 from the variation with respect to @xmath13 and @xmath36 , we obtain the equations of motion , @xmath37 where we also used the constraint equation ( [ frweq ] ) to write the equation in the form as given in ( [ doth ] ) .
we refer to the above two equations , ( [ feq ] ) and ( [ doth ] ) , as the background equations . requiring the regularity at both boundaries , the boundary conditions to be satisfied by the background solutions
are determined as @xmath38 recently , hawking and turok proposed to relax these boundary conditions to allow singular instantons@xcite . however , the interpretation of singular instantons is still an unsettled issue@xcite .
here we do not consider this possibility .
now , we consider fluctuations around this bounce solution , to examine the prefactor @xmath11 that arises in the estimate of the decay rate . as
briefly noted in introduction , the number of @xmath10 in the prefactor @xmath11 is important in the path integral approach . to evaluate the number of @xmath10
, we need to obtain the reduced action for the fluctuations around the bounce solution .
especially , @xmath12-symmetric perturbations @xmath39 are most important .
after an appropriate gauge fixing , we obtain the reduced action for @xmath12-symmetric perturbations as@xcite @xmath40 , \label{action2}\ ] ] with the potential @xmath41 , @xmath42 where @xmath43 is the @xmath12-symmetric part of the gauge invariant variable introduced in ref .
@xcite , and @xmath44 is its conjugate momentum . the euclidean conformal time @xmath45 is related to @xmath26 by @xmath46 . as given in ref .
@xcite , perturbations in terms of original variables are written down in the newton gauge as @xmath47 where @xmath48 denotes the differentiation with respect to @xmath45 .
a distinguishable feature of the reduced action for @xmath12-symmetric perturbations is that the coefficient in front of @xmath49 becomes negative .
note that the prefactor @xmath11 is evaluated by @xmath50e^{-\delta^{(2 ) } s}.\ ] ] thus , in doing the integration with respect to momentum variables , we will find that the gaussian integral does not converge . to resolve this difficulty
, we proposed to use the prescription similar to the conformal rotation@xcite . by changing the variables like @xmath51 , @xmath52 ,
the above integration becomes well - defined .
if we discretize the path integral , the numbers of @xmath44 and @xmath43 integrations are different by 1 . therefore , this change of variable will produce one imaginary unit @xmath10 .
that is , @xmath53e^{\delta^{(2 ) } s } \cr & \approx & i\int [ dq ] \exp\left[- { 3\over 2}\int d\eta
\left\{{1\over 2}\left({dq\over d\eta}\right)^2 + { 1\over 2}(u-3 ) q^2\right\ } \right ] .
\hspace*{-2 cm } \cr & & \end{aligned}\ ] ] to determine the number of @xmath10 in @xmath11 , we also need to know the spectrum of eigen values , @xmath54 , of the following schrdinger type equation : @xmath55 the contribution to @xmath11 from @xmath12-symmetric perturbations will be given by @xmath56 .
if there is no mode which has a negative eigen value , there arises no additional imaginary unit factor .
then , the prefactor @xmath11 becomes imaginary .
thus , we conjectured that there should not be any negative mode for the system including the gravitational degrees of freedom , in ref
. @xcite . as discussed in ref .
@xcite , one can show that , if there is no - negative mode in @xmath12-symmetric perturbations , it is also the case for the other perturbation modes . hence , it is sufficient to concentrate on @xmath12-symmetric perturbations . as we have shown in ref .
@xcite , we can construct models of cd bounce solutions which possess negative modes .
hence , we can not prove the absence of negative modes for arbitrary models of cd bounce solutions .
however , as mentioned in introduction , the important point is that not all cd bounce solutions contribute to the tunneling process .
the bounce solution that determines the decay rate is the one that realizes the minimum value of action among all bounce solutions . here , we strictly define the `` no - negative mode theorem '' , a proof of which we are going to present in this paper .
* the main theorem * ( the no - negative mode theorem ) :
we develop a method to construct a complete list of @xmath12-symmetric cd bounce solutions in this section . for this purpose
, we introduce a diagram consisting of two curves .
we show that each intersection point of these curves corresponds to a cd bounce solution . in solving the background equations , if we fix the boundary value of @xmath13 at one side , say @xmath57 at @xmath58 , the boundary conditions ( [ boundary ] ) completely determines the initial condition to solve the equations of motion , ( [ feq ] ) and ( [ doth ] ) .
then , solving eqs .
( [ feq ] ) and ( [ doth ] ) from both sides to the maximum radius , we obtain functions which satisfy the background equations in the respective half euclidean regions . we denote them by @xmath59 where the subscript @xmath60 is attached with the variables that are used in solving the background equations from the true ( false ) vacuum side .
when we solve the background equations from @xmath61 , the initial condition is varied by sweeping @xmath62 between the true vacuum minimum and the top of the potential , i.e. , @xmath63 . while , when we solve them from @xmath64 , @xmath65 is sweeped between the false vacuum minimum and the top of the potential , i.e. , @xmath66 . here , we move the origin of the @xmath26-coordinate to the point corresponding to the maximum radius , i.e. , @xmath67 .
thus , @xmath28 is determined by solving @xmath68 , and becomes a function of @xmath69 .
further , we define two functions of @xmath69 by the values of @xmath70 and @xmath71 at @xmath72
. we refer to them as @xmath73 and @xmath74 , respectively , where the subscript @xmath75 is used in the same manner as before
. then , as shown in fig.[fig2]@xmath76 , the points @xmath77 @xmath78 and @xmath79 @xmath80 draw two curves , @xmath81 and @xmath82 , on the @xmath83-plane as @xmath84 and @xmath85 are sweeped .
since @xmath86 stays at @xmath87 when @xmath88 or @xmath89 , the curve @xmath90 starts from @xmath91 and terminates at @xmath92 and the curve @xmath81 starts from @xmath93 and terminates at @xmath94 .
we refer to the curves moving in the opposite direction as @xmath95 . : _ each of the curves @xmath96 does not intersect with itself_. and @xmath97 by connecting points @xmath98 and @xmath99 , respectively . by the same reasoning
, we can also prove that @xmath100 does not intersect with @xmath101 . ]
if it were the case , we would have the same final values of @xmath13 and @xmath102 at @xmath72 for different solutions .
however , the background equations can be solved reversely from @xmath72 , and the evolution is uniquely determined .
this is a contradiction .
@xmath103 : _ all intersection points between @xmath104 and @xmath90 correspond to cd bounce solutions . @xmath105an exceptional case is the hawking moss instanton corresponding to @xmath106 _ we set @xmath107 for @xmath108 and @xmath109 for @xmath110 .
then , the values and the first derivatives of @xmath111 are continuous at @xmath72 for intersection points . with the aid of eq.([frweq ] ) we find that @xmath112 is also continuous at @xmath72 .
since @xmath113 and @xmath114 , respectively , satisfy the regular boundary conditions at @xmath58 and at @xmath115 , @xmath116 satisfies those at both boundaries .
hence , @xmath116 is a cd bounce solution .
@xmath103 before closing this section ,
we show that there is another general constraint on @xmath96 .
we define a function @xmath117 where we used eq.([frweq ] ) with @xmath118 . as a function of @xmath119 and @xmath120 ,
the definition of @xmath121 is implicit . to determine @xmath121 from this equation
, we need to solve a third order algebraic equation like @xmath122 , where @xmath123 and @xmath124 are real .
the left hand side of this equation is a monotonic function of @xmath121 , and hence the real solution for @xmath121 is unique .
we schematically plot the @xmath121-constant contours in fig.[fig2]@xmath125 . from eq .
( [ feq ] ) , we can see that @xmath126 decreases monotonically in the direction for @xmath112 to increase .
thus , @xmath127 is bounded by @xmath128 * proposition 3 * : _ the curves @xmath96 do not enter the shaded regions labeled with @xmath129 and @xmath130 in fig.[fig2]@xmath125 .
_ it is manifest that @xmath90 can not go into the regions @xmath129 and @xmath130 in fig.[fig2]@xmath125 from the bound ( [ rhobound ] ) . on the other hand , from ( [ rhobound ] ) alone , it seems possible that the curve @xmath104 enters into the region @xmath130 .
however , for the potential that we assumed , @xmath131 stays negative once @xmath132 becomes smaller than @xmath19 .
hence , when we solve equations of motion from @xmath58 , @xmath132 never stops on the left side of @xmath19 .
therefore , when @xmath133 , @xmath134 must be positive .
@xmath103 for the later convenience , we introduce a new coordinate , @xmath135 , where @xmath121 is given by ( [ rhodef ] ) and @xmath136 is defined by @xmath137 for the unshaded region in fig.[fig2]@xmath125 . for the region @xmath138 , which the curve @xmath104 can enter
, we extend the @xmath136-coordinate so that it change monotonically along each @xmath121-constant curve and does not have a point with @xmath139 in this region , where @xmath140 is an integer .
in this section , we develop a method to count the number of negative modes of eq .
( [ yeq ] ) from the topology of the @xmath83-diagram introduced in the preceding section . unless @xmath141 goes to @xmath22 , eq .
( [ yeq ] ) is a schrdinger type equation with a regular potential . in this case
, the number of nodes of @xmath142 for @xmath143 gives the number of negative modes .
hence , as long as such nodeless bounce solutions are concerned , we can say that there is no - negative mode when @xmath142 for @xmath143 has no node .
therefore , we concentrate on the case with @xmath143 .
although the above statement does not hold any longer once a node of @xmath141 appears , we do not assume @xmath144 in most of the following discussions in this section .
the only exception is the subsection c. in subsection a , we introduce variables @xmath145 defined by eq.([qeq ] ) with eq.([varphi ] ) .
they are constructed from the background solution @xmath146 , and we show that they satisfy eq .
( [ yeq ] ) for @xmath143 with @xmath147 .
then , in subsection b , we give a method to count the number of nodes of the functions @xmath148 from the @xmath83-diagram .
further , we define a topological number @xmath34 assigned for each intersection point , and show that it should be non - negative . in subsection c , we prove the following theorem . :
we begin this subsection by pointing out that eq .
( [ yeq ] ) for @xmath143 is nothing but the perturbed equations of motion that are obtained by taking the variation of the reduced action . since the @xmath12-symmetry is the symmetry that the background solution possesses , it is natural to consider that the case with @xmath143 is related to the background solutions
: _ we define functions @xmath149 and @xmath150 by @xmath151 where @xmath152 then , @xmath145 satisfies eq.@xmath153 for @xmath143 with @xmath154 regular boundary conditions . _
first , we derive the equation satisfied by @xmath155 .
hereafter , we sometimes suppress the subscript @xmath75 to keep the simplicity of notion .
it is obtained by differentiating the background equations with respect to @xmath156 . from eq.([feq ] ) , we have @xmath157 from eqs.([frweq ] ) and ( [ doth ] ) , we have @xmath158 differentiating this equation with respect to @xmath156 , we obtain @xmath159 by eliminating @xmath160 from eqs .
( [ dphi1 ] ) and ( [ dphi2 ] ) , we obtain a third - order differential equation with respect to @xmath161 . after a lengthy but straightforward calculation
, one can verify that the equation that @xmath161 satisfies reduces to @xmath162 where we used the relation @xmath163 form eq.([eqvarphi ] ) , it is manifest that @xmath164 defined by ( [ qeq ] ) satisfies eq .
( [ yeq ] ) for @xmath143 with @xmath147 .
we note that @xmath164 is defined not only for intersection points corresponding to cd bounce solutions but also for any points on the curves @xmath96 .
next , we show that @xmath164 satisfies the required boundary condition , i.e. , @xmath165 for @xmath166 . from the boundary condition for the background solutions , we have @xmath167 in the @xmath166 limit , where @xmath168 and @xmath169 are positive constants depending on @xmath87 . by substituting these expressions into eq .
( [ qeq ] ) , we can show that @xmath170 thus , @xmath149 and @xmath150 satisfy the appropriate boundary condition .
@xmath103 as a bonus , we also find that @xmath171 here , we show that the expression ( [ qeq ] ) can also be derived by considering the relation between the gauge invariant variable @xmath43 and @xmath172 in the synchronous gauge , which we refer to as @xmath173 . here
we note that we used the synchronous gauge in solving the background equations . considering the case in which @xmath43 satisfies eq.([yeq ] ) for @xmath143 , i.e. , @xmath174
, we give a gauge invariant representation of @xmath43 in terms of the original variables .
since the gauge transformation @xmath175 acts on @xmath12-symmetric perturbations as @xmath176 we find that @xmath177 is a gauge invariant combination . by using the expressions in the newton gauge ( [ newtong ] ) , we can verify that this is the gauge invariant expression of @xmath43 in terms of the original variables .
henceforce , we finally find that @xmath43 is related to @xmath178 as @xmath179 we develop a method to count the number of nodes of @xmath181 , where @xmath164 is defined by eq .
( [ qeq ] ) . for the convenience ,
we use @xmath180 instead of @xmath164 , but the number of nodes of @xmath180 equals to that of @xmath164 if @xmath182 stays positive . : _ as @xmath87 is varied along @xmath183 , the number of nodes of @xmath184 @xmath105existing between @xmath22 and @xmath28 excluding the point @xmath185 changes only when @xmath186 .
_ it is sufficient to prove that the location of a zero point of @xmath184 should change continuously when @xmath87 is sweeped . to show this , we suppose the opposite case .
if a zero point of @xmath184 suddenly appeared at a point @xmath187 other than @xmath72 , @xmath188 would have to vanish there due to the continuity of @xmath184 with respect to @xmath87 .
except for the case that @xmath182 vanishes at @xmath187 , the conditions that @xmath189 imply @xmath190 , and hence @xmath191 because @xmath164 satisfies a regular second order differential equation .
hence , the remaining possibility is that @xmath182 vanishes at @xmath187 .
then , let us examine the power series solution of @xmath164 around the singular point @xmath192 . near the singularity
, the behavior of @xmath193 is given by @xmath194 , where @xmath195 is the value of @xmath45 corresponding to @xmath187 , and @xmath196 is a constant .
then , the potential @xmath41 with @xmath147 in eq.([yeq ] ) behaves as @xmath197 .
hence , the point @xmath198 is a regular singular point , and the exponents are @xmath199 and @xmath200 .
thus , we find that the general power series solution for @xmath180 around this singular point is given by @xmath201 , where @xmath202 and @xmath203 are constants .
since the case @xmath204 is considered , @xmath202 must be zero .
therefore , @xmath180 must change its signature at @xmath187 .
hence @xmath187 can not be the first point at which the zero point suddenly appeared .
@xmath103 to represent the condition @xmath205 in terms of the directions of @xmath96 , @xmath206 we first calculate @xmath207 as @xmath208 where we used eq.([frweq ] ) with @xmath118 . from this equation
, we can express @xmath209 as a function of @xmath210 and @xmath207 . substituting this into eq .
( [ qeq ] ) , we obtain @xmath211 by using eq .
( [ rhodef ] ) , we find the above equation can be written as @xmath212 for the convenience , using the @xmath213-coordinate introduced at the end of sec.[oldsec2 ] , we define angles @xmath214 and @xmath215 by @xmath216 along @xmath217 and @xmath82 , respectively . here , @xmath218 and @xmath219 represent points corresponding to @xmath220 and @xmath221 @xmath222 in the @xmath213-coordinate , respectively . to remove the ambiguity in @xmath223 modulo @xmath224 , we set @xmath225 and @xmath226 .
when the curves @xmath90 and @xmath227 start from the vacua , both @xmath228 and @xmath229 have no node , the proof of which is presented in the appendix . furthermore , in the @xmath166 limit , it is easy to see @xmath230 , and @xmath231 was shown in eq .
( [ mpq ] ) . combining these facts , we find that both @xmath232 and @xmath233 are also positive when the curves @xmath90 and @xmath227 first start from the vacua
. then @xmath214 and @xmath215 are defined without any ambiguity by choosing them to be continuous along the curves @xmath234 with the condition that @xmath235 in terms of @xmath236 , the point at which the number of nodes of @xmath164 changes is specified by @xmath237 , where @xmath238 is an integer . to determine whether the number of nodes increases or decreases at a point with @xmath237 , we examine the signature of @xmath239 .
it is calculated from eq .
( [ qeq ] ) with the aid of eqs.([feq ] ) and ( [ dphi1 ] ) as @xmath240 where we used @xmath241 .
then , from fig.[fig2]@xmath125 , we find @xmath239 is positive when the curve @xmath82 or @xmath227 touches a @xmath121-constant curve with its direction pointing in the @xmath136-direction . by comparing this with eq.([thetadef ] ) , the defining eqaution of @xmath236 , we find that @xmath239 is positive when @xmath242 , while @xmath239 is negative when @xmath243 , where @xmath238 is an integer . to summarize , we have @xmath244 [ fig3 ] : _ we refer to the number of nodes of @xmath184 as @xmath245
. then @xmath245 is determined by _
@xmath246 _ here we do not count a zero point at @xmath72 as a node even if it exists . _
let us consider the @xmath108 side .
when @xmath247 , @xmath248 changes its signature from @xmath249 to @xmath250 for increasing @xmath215 .
remember that @xmath251 in this case .
then , it will be easy to see from fig.[fig3 ] that the number of nodes , @xmath252 , increases by one when @xmath215 crosses @xmath253 in its increasing direction .
when @xmath254 , @xmath255 is negative . in this case ,
@xmath256 changes its signature from @xmath250 to @xmath249 for increasing @xmath215 .
then , again the number of nodes , @xmath252 , increases by one when @xmath215 crosses @xmath257 in the increasing direction . for the @xmath110 side
, we can also give an analogous discussion .
using the fact that @xmath164 is nodeless when the curves @xmath258 start from @xmath259 with @xmath260 ( see appendix ) , the proof is completed .
@xmath103 now , we are ready to assign a number @xmath34 for each intersection point .
we define @xmath34 by the relation , @xmath261 since both @xmath262 and @xmath252 must be positive , we have @xmath263 .
thus we obtain : @xmath264 in this subsection , we give a proof of theorem 1 .
we assume @xmath102 stays positive throughout this subsection .
first , we explain the fact that the number of negative modes at an intersection point @xmath265 is not simply given by the summation of numbers of nodes of @xmath164 , i.e. , @xmath266 . in order to obtain the number of negative modes , the information about @xmath262 and @xmath252 must be supplemented with the knowledge of the signature of @xmath267 * proposition 8 * : _ the number of negative modes is given by _
@xmath268 we define @xmath269 by the solution of eq .
( [ yeq ] ) that satisfies regular boundary conditions imposed at one boundary @xmath270 .
then , the quantity corresponding to @xmath271 can also be defined for @xmath272 by replacing @xmath273 with @xmath269 in ( [ defb ] ) .
we denote it by @xmath274 . in the present case ,
the potential @xmath41 is non - singular .
then , if we gradually decrease @xmath54 , @xmath274 also decreases monotonically until it diverges to @xmath275 .
when @xmath274 diverges , either @xmath276 or @xmath277 vanishes , and hence the number of nodes changes . in the case of @xmath278 ,
decreasing @xmath54 from zero , @xmath274 vanishes at some @xmath54 before the number of nodes of @xmath279 changes .
this @xmath54 is the largest negative eigen value , and the corresponding eigen function still has @xmath266 nodes .
hence , the number of negative modes is @xmath280 . on the other hand , when @xmath281 , we gradually increase @xmath54 .
then , we have the first eigen mode with positive @xmath54 which has @xmath266 nodes .
hence , the number of negative modes is @xmath266 .
@xmath103 [ fig4 ] from proposition 8 , we find that to show theorem 1 is equivalent to show the following proposition .
: we first consider which case is realized for given @xmath236 .
for example , we fix @xmath215 as shown in fig.[fig4 ] , which is a close - up view around an intersection point in the @xmath83-diagram as shown in fig.[fig2]@xmath76 .
this figure is rotated so as for the @xmath136-direction to point rightward .
then , we classify the possible direction of the curve @xmath282 into four regions labeled with @xmath129 , @xmath130 , @xmath283 and @xmath284 .
the boundaries of four regions belong to the region with the smaller @xmath214 .
then , we can see that case ( a ) , in which @xmath285 , is realized when @xmath286 is in the region @xmath129 or @xmath283 .
to show @xmath287 in this case , we evaluate @xmath271 by using eq .
( [ pim ] ) to find @xmath288 in the regions @xmath129 and @xmath130 , the quantity in the round bracket is positive . on the other hand ,
the quantity @xmath289 is positive in the regions @xmath129 and @xmath284 .
consequently , we obtain @xmath287 in the regions @xmath129 and @xmath283 .
the case in which the curve @xmath90 is pointing upward can be discussed in the same way .
@xmath103 : the value of @xmath290 at @xmath265 does not cross @xmath291 under a continuous deformation of the curves @xmath96 as long as the way of intersection at @xmath265 is unchanged and the curves do not cross the points @xmath93 , @xmath292 and @xmath293 . hence , the number @xmath34 at a point @xmath265 is invariant under such a continuous deformation of @xmath96 .
from our method to find solutions from the intersection points of two curves @xmath96 , the solution may have several nodes of @xmath102 .
we show this number of nodes , @xmath294 , can also be read from the @xmath83-diagram . for this purpose , we prove the following statement . : _ a point where @xmath295 appears only from the boundary specified by @xmath72 , as we vary @xmath87 along the curves @xmath96 .
_ let us assume that a point at which @xmath296 appeared at an intermediate point @xmath297 abruptly .
then , we must have @xmath298 from the continuity of @xmath299 with respect to both @xmath87 and @xmath26 .
however , the field equation ( [ feq ] ) indicates that @xmath300 at this point .
then , this solution must be one of the trivial solutions @xmath301 or @xmath302 , which correspond not to an intermediate point of @xmath96 but to their end points .
this is a contradiction . @xmath103
* theorem 2 * : @xmath303 _ proof .
_ from proposition 9 , the number of nodes of @xmath304 changes only when @xmath305 as we change @xmath87 along the curves @xmath96 .
this happens if and only if @xmath306 where @xmath140 is an integer .
denote by @xmath307 the number of nodes of @xmath182 @xmath105existing between @xmath22 and @xmath28 excluding the point @xmath185 . to avoid duplicative counting
, we do not count the point @xmath72 as a node when @xmath308 but we count it as a node when @xmath309 . here
we use the same technique that was used to count the number of nodes of @xmath180 in the subsection [ subn2 ] .
corresponding to eq.([xbardot ] ) , the derivative of @xmath182 with respect to @xmath26 at @xmath72 is calculated as @xmath310 this quantity is positive ( negative ) when @xmath311 is negative ( positive ) . then , for the same reasoning as was used to determine the number of nodes for @xmath145 , we find the relations , @xmath312 for an intersection point , @xmath111 is given by @xmath313 for @xmath108 and @xmath314 for @xmath315 .
thus , the number m , which is the total number of nodes of @xmath316 , is given by @xmath317 .
then , from eq.([mbound ] ) , we find @xmath318 @xmath319 @xmath320 . since @xmath321 must be devided by @xmath224 , we finally find @xmath294 is given by ( [ numm ] ) .
@xmath103 : theorem 2 tells that @xmath322 is the winding number around @xmath93 of the continuous curve @xmath323 that is obtained by connecting @xmath324 and @xmath292 with @xmath90 and @xmath104 .
thus , the number @xmath294 is invariant under a continuous deformation of curves @xmath96 as long as the curves do not cross the points @xmath93 , @xmath292 and @xmath293 .
we introduce a pseudo - action defined as a function of @xmath325 and @xmath326 by @xmath327 and @xmath328,\ ] ] where @xmath329= a^3 \left[{1\over 2}\dot\phi^2+v(\phi ) -{3\over \kappa } \left(\left({\dot a\over a}\right)^2 + { 1\over a^2}\right)\right ] .
\label{partaction}\ ] ] note that @xmath330 is defined along the curve @xmath96 , and hence so @xmath331 is . to the contrary ,
the original action @xmath32 is well - defined only at intersection points which represent cd bounce solutions . for cd bounce solutions ,
i.e. , at intersection points , we have @xmath332 .
then , we calculate @xmath333\over \delta a_{\pm } } { \partial a_{\pm}\over \partial\phi_{\pm}^i } \cr & & \quad\quad\quad\quad\quad + { \delta l[a_{\pm},\phi_{\pm}]\over \delta \phi_{\pm } } { \partial \phi_{\pm}\over \partial\phi_{\pm}^i}\bigr ) \cr & & \quad\quad\quad + \left [ -{6\over \kappa } a_{\pm}\dot a_{\pm } { \partial a_{\pm}\over \partial\phi_{\pm}^i } + a^3_{\pm}\dot\phi_{\pm}{\partial \phi_{\pm}\over \partial\phi_{\pm}^i } \right]_{0}^{\tau_{\pm } } \cr & & \quad\quad\quad + { d\tau_{\pm}(\phi^i)\over d\phi_{\pm}^i } l[a_{\pm},\phi_{\pm}]|_{\tau=\tau_{\pm } } \biggr\ } \cr & = & \mp 2\pi^2 \bar\pi_{\pm } \bar\varphi_{\pm } , \end{aligned}\ ] ] where partial differentiations with respect to @xmath87 are taken for fixed @xmath26 , and we used the fact @xmath334 satisfies the background equations , i.e. , @xmath335 integrating this expression , we find that the difference of the values of action between two bounce solutions specified by points @xmath336 and @xmath337 is given by @xmath338-s_e[p_1]=2\pi^2 \int_{p_1}^{p_2 } d\bar\phi ( \bar\pi_{- } -\bar\pi_{+}).\ ] ] when the two curves connecting @xmath336 and @xmath337 do not intersect with each other , the signature of @xmath339-s_e[p_1]$ ] is totally determined by the topological information of [ fig6 ] the connecting curves . as shown in fig.[fig5]@xmath76 ,
if the region surrounded by the two curves is on the left hand side when we move from @xmath336 to @xmath337 along the curve @xmath90 , @xmath340>s_e[p_2]$ ] can be concluded .
the difference is proportional to the area surrounded by the two curves . in the opposite case as shown in fig.[fig5]@xmath125
, we can say @xmath340<s_e[p_2]$ ] .
in such a case as presented in fig.[fig5]@xmath341 , we need to compare the areas @xmath342 and @xmath343 to determine which point has the smaller value of action . in some cases ,
however , even when there is another intersection point between @xmath336 and @xmath337 as shown in fig.[fig5]@xmath344 , we can determine which point has the smaller value of action from the topological information alone .
in the case of fig.[fig5]@xmath344 , we can say that @xmath340<s_e[p_3]$ ] and @xmath345<s_e[p_2]$ ] .
hence @xmath340<s_e[p_2]$ ] is concluded . for our present purpose ,
we do not have to give a general statement in what situation we can determine the signature of @xmath339-s_e[p_1]$ ] from the topological information alone . only the case discussed below is of special importance .
let us draw the two curves @xmath96 by using a spherical coordinate @xmath346 whose origin is located on @xmath337 , i.e. , the point @xmath337 corresponds to the line @xmath347 .
in this coordinate , when the value of @xmath348 of two points are different by @xmath253 with an integer @xmath238 , these two points are identical on the original @xmath83-plane . thus , there are many copies of the curves @xmath96 in the @xmath346-coordinate .
we choose one pair of @xmath96 both of which pass through a common point corresponding to @xmath336 in the @xmath346-coordinate , and denote them by @xmath349 .
if we choose this point , say , to satisfy @xmath336 with @xmath350 , the two curves @xmath351 can be drawn without any ambiguity .
if the two curves @xmath349 do not intersect between @xmath336 and @xmath337 in this coordinate as shown in fig.[fig6 ] , where we used @xmath352 as the horizontal coordinate instead of @xmath353 , the @xmath354-s_e[p_1])/2\pi^2 $ ] is given by the area of the shaded region , and we can say it is positive . we note that no intersection in this coordinate does not mean no intersection on the original @xmath83-plane . in fact , fig.[fig6 ] represents the same situation that was shown in fig.[fig5]@xmath344 .
the result obtained in this section is summarized by the following theorem .
: : theorem 3 depends only on the topological information about intersections between the curves @xmath104 and @xmath90 .
hence , we can apply the same statement for any deformed diagram which preserves this topological information .
now , we prove the main theorem ( the `` no - negative mode theorem '' ) .
namely , we prove if an intersection point @xmath265 has either negative modes or nodes of @xmath316 , there exists another intersection point @xmath355 which has a smaller value of action than @xmath265 . by using theorem 1 given in sec.[oldsec5 ] and theorem 2 given in sec.[oldsec6 ]
, the main theorem is reduced to the following statement . : to show the above statement , we first define curves @xmath356 by deforming @xmath96 continuously to make @xmath82 to be a straight line connecting @xmath293 and @xmath93 . under this deformation of curves , the way of intersections , i.e. , the value of @xmath357 at each intersection point , is kept unaltered , and the curves do not cross the points @xmath93 , @xmath292 and @xmath293 .
the diagram composed of these deformed curves keeps the numbers @xmath34 and @xmath294 assigned for each intersection point , and also it still keeps some information about the difference between the values of action for different points .
in fact , under the situation such that theorem 3 determines which point has the smaller value of action , we can apply the theorem to this deformed diagram in place of the original one .
we denote this diagram consisting of @xmath356 by @xmath358 . focusing on a point @xmath265 in @xmath358
, we further deform @xmath359 continuously . under this deformation
, we allow the way of connections at points other than @xmath265 to change , and hence the number of intersection points can be varies .
furthermore , we allow the value of @xmath290 at @xmath265 to deviate from @xmath360 , where @xmath360 is the original value of @xmath290 at @xmath265
. however , we do not allow the curve to cross the points @xmath93 , @xmath292 and @xmath293 .
hence , the number @xmath294 for @xmath265 is kept unaltered .
then , the diagram can be reduced to the diagram that are shown in fig.[fig7]@xmath76 .
the cases given by fig.[fig7]@xmath125 are not realized because the number @xmath294 for the point @xmath265 becomes negative .
we denote these further deformed curves by @xmath361 ( c.f . , @xmath362 and the corresponding diagram by @xmath363 .
as noted in the previous section , the number @xmath294 represents twice the number of windings of the curve @xmath323 around @xmath93 .
now we use the @xmath364-coordinate introduced in sec.[oldsec3 ] . in this coordinate ,
the curves have many copies .
as before , we attach @xmath365 to a specific pair of curves to distinguish it from others .
it is explained below how we select the specific curves .
the deformed diagram @xmath363 in this coordinate is given by fig.[fig8]@xmath76 when @xmath366 and by fig.[fig8]@xmath125 when @xmath367 .
in both cases , @xmath368 and @xmath369 are shown by the thick rigid line and by the thick dashed line , respectively . here ,
@xmath370 in fig.[fig8]@xmath76 is the intersection point neighboring to @xmath265 in fig.[fig7 ] @xmath76 .
later , we consider the process that the diagram @xmath358 is recovered from @xmath363 through continuous deformation of curves . in the midst of this process
, we use the notation @xmath349 to specify the curves that are denoted by @xmath371 in the diagram @xmath363 .
we refer to the point corresponding to the point @xmath265 on the curve @xmath372 by @xmath373 , ( do not be confused with @xmath292 ! ) and we denote the value of @xmath348 at @xmath373 by @xmath374 . in both case @xmath76 and case @xmath125 , we choose @xmath375 so as to satisfy @xmath376 .
the part of the curve @xmath377 connecting @xmath293 and @xmath265 is given by the lines with @xmath378 , where @xmath238 is an integer .
we choose the line with @xmath379 as @xmath380 for case @xmath76 , and that with @xmath381 for case @xmath125 .
they are presented by the thick dashed lines in fig.[fig8 ] .
: _ we consider the diagram @xmath358 that is recovered from the diagram @xmath363 given in fig.[fig8]@xmath76 or @xmath125 .
for this diagram , @xmath382 $ ] is given by @xmath383 $ ] , where @xmath360 is the value of @xmath290 at @xmath265 in the original diagram .
@xmath105here we denoted the largest integer less than or equal to @xmath384 by @xmath385.)$ ] furthermore , @xmath386 in the diagram @xmath358 must be greater than @xmath22 . _
_ proof . _ from fig.[fig7]@xmath76 , one can see that the value of @xmath290 at @xmath265 for this deformed diagram @xmath363 is also in between @xmath22 and @xmath387 .
hence , we find @xmath388=[\theta/\pi]$ ] for the diagram @xmath363 . in recovering the diagram @xmath358 from @xmath363 ,
the direction of the curve @xmath389 in fig.[fig7]@xmath76 may rotate around @xmath265 . under this rewinding ,
both @xmath290 and @xmath374 increase when the direction of @xmath104 rotates in the the anti - clockwise direction , and both of them are divided by @xmath387 when @xmath104 becomes parallel to @xmath90 at @xmath265 .
hence , we can say that the relation @xmath390=[\theta_+/\pi]$ ] is maintained under the deformation from @xmath363 to @xmath358 . therefore , when the diagram @xmath358 is recovered , the point @xmath373 is moved along the @xmath348-axis to a point satisfying @xmath391=[\theta_{(p)}/\pi]$ ] .
the latter part of the proposition is manifest from proposition 7 , which tells @xmath360 must be positive . @xmath103
* theorem 4 * : let us consider the continuous deformation of curves to recover the diagram @xmath358 from @xmath363 in the @xmath364-coordinate . as shown in fig.[fig8]@xmath76 , @xmath368 has an intersection point @xmath370 with @xmath392 . from the continuity of deformation , the curve @xmath393 , as well as @xmath368 , must intersect with @xmath394 . then , let us refer to the first intersection point nearest to @xmath265 as @xmath355 .
since the curve @xmath395 does not intersect with the curve @xmath394 between @xmath265 and @xmath355 by definition , the area surrounded by these two curves connecting @xmath265 and @xmath355 has a definite signature . by virtue of proposition 10 ,
@xmath373 in the diagram @xmath358 must be on the upper side of the @xmath396 line .
hence , by using theorem 3 , we conclude @xmath397 < s_e[p]$ ] .
@xmath103 * theorem 5 * : again , we consider the continuous deformation to recover the diagram @xmath358 from @xmath363 in the @xmath364-coordinate . in the case
@xmath367 , the diagram @xmath363 is given by fig.[fig8]@xmath125 .
if @xmath398 , @xmath360 must be greater than @xmath387 .
then , owing to proposition 10 , @xmath374 for the diagram @xmath358 must be greater than @xmath387 . consequently , we find that the point @xmath373 moves to the upper side of the @xmath399 line , while the point @xmath400 must stay on the opposite side .
recalling the forbidden region shown as shaded regions in fig.[fig8]@xmath125 , we find that the curve @xmath401 must have at least one intersection point with @xmath402 .
let us refer to the first intersection point nearest to @xmath265 as @xmath355 .
then , as before , we can use theorem 3 to conclude @xmath397 < s_e[p]$ ] .
@xmath103 now , from theorem 4 and theorem 5 , the proof of the main theorem is completed .
we gave a proof of the `` no - negative mode theorem '' , which tells that the bounce solution realizing the smallest value of action among all @xmath12-symmetric cd bounce solutions has no - negative mode and has no node in @xmath102 .
we summarize the outline of the proof presented in this paper . in sec.iii
, we introduced a diagram which consists of two curves .
we showed that each intersection point of these two curves corresponds to an @xmath12-symmetric cd bounce solution . in sec.iv and v ,
we assigned two numbers @xmath34 and @xmath294 for each intersection point .
the number @xmath34 is defined by ( [ numn ] ) , and it was shown to represent the number of negative modes when @xmath367 .
the number @xmath294 is defined by ( [ numm ] ) , and it was shown to be the number of nodes of @xmath102 . as is easily seen from their definitions , both @xmath34 and @xmath294 have a clear topological meaning . in sec vi
, we gave a general rule to compare the values of action between different intersection points from the topological information of the diagram introduced in sec.[oldsec2 ] .
collecting these statements related to the topology of the diagram , the `` no - negative mode theorem '' was proved in sec.vii .
the `` no - negative mode theorem '' is known to be equivalent to the `` no - supercritical super curvature mode conjecture''@xcite .
therefore , we also proved that there appears no - supercritical supercurvature mode in the one - bubble open inflation universe . in this appendix , we show that @xmath403 is positive in the @xmath404 limit . for @xmath405
, @xmath406 will stay near @xmath407 , and the geometry does not significantly differ from the de sitter space . in both @xmath250 and @xmath249 cases ,
the limiting behavior of @xmath403 is essentially same . to avoid an unnecessary complication
, we discuss the false vacuum side as a representative case .
similar arguments can be repeated for the true vacuum side .
the equation for @xmath408 can be approximately written as @xmath409 where @xmath410 . since @xmath411 is almost independent of @xmath412 , eq .
( [ deltaeq ] ) reduces to a linear differential equation .
hence , in this limiting case , solutions of eq .
( [ deltaeq ] ) with different initial values of @xmath13 are obtained by a simple scaling . by introducing a scaling parameter @xmath413 ,
@xmath412 is written as @xmath414 where @xmath415 satisfies eq.([deltaeq ] ) .
then , we have @xmath416 where @xmath417 is positive , and hence @xmath418 because @xmath419 . substituting these approximate expressions into ( [ qeq ] ) ,
we find @xmath420 where @xmath421 .
since @xmath422 and @xmath423 stay non - negative in the present limiting case , we have @xmath424 .
furthermore , the equation that @xmath425 should satisfy is given by @xmath426 at @xmath166 , @xmath427 .
then if @xmath425 became greater than @xmath140 , there would be a point at which both @xmath428 and @xmath429 are satisfied .
however this is impossible from eq .
( [ omegadot ] ) .
thus , @xmath430 is maintained .
then , the condition @xmath431 with eq.([qmd ] ) implies that @xmath432 stays positive until @xmath72 .
gott iii , nature * 295 * , 304 ( 1982 ) ; j.r .
gott iii and t.s .
statler , phys .
* 136b * , 157 ( 1984 ) ; m. sasaki , t. tanaka , k. yamamoto and j. yokoyama , phys .
b * 317 * , 510 ( 1993 ) ; m. bucher , a.s . goldhaber and n. turok ,
phys . rev .
* d52 * 3314 ( 1995 ) ; m. bucher and n. turok , phys . rev . *
d52 * , 5538 ( 1995 ) ; a. linde .
b351,99 ( 1995 ) ; a. linde and a. mezhlumian , phys .
* d52 * , 6789 ( 1995 ) ; j. garcia - bellido , j. garriga and x. montes , phys . rev . * d57 * , 4669 ( 1998 ) . m. sasaki , t. tanaka and k.yamamoto , phys . rev .
* d51 * , 2979 ( 1995 ) ; j. garcia - bellido , a.r .
lidle , d.h .
lyth and d. wands , phys .
rev . * d55 * , 4596 ( 1997 ) .
t. tanaka and m. sasaki , phys . rev .
* d * ( 1998 ) _ in press_. t. tanaka and m. sasaki , prog .
* 88 * ( 1992 ) , 503 .
s. coleman , phys . rev . *
d15 * , ( 1977 ) , 2929 ; in _ the why s of subnuclear physics _ ,
proceedings of the international school , erice , italy , ed .
a. zichichi , subnuclear series vol.15 ( plenum , new york . 1979 ) , p.805 . c. g. callan , jr . and s. coleman , phys . rev .*d16 * , 1762 ( 1977 ) .
s. coleman , nucl .
b298 * , 178 ( 1988 ) ; s. coleman , v. glaser and a. martin , comm .
* 58 * ( 1978 ) , 211 s. coleman and f. de luccia , phys . rev . *
d21 * , 3305 ( 1980 ) .
dirac , lectures on quantum mechanics , yeshiva university , ( 1964 ) .
j. garriga , x. montes , m. sasaki and t. tanaka , nucl .
b*513 * , 343 ( 1998 ) .
s.w . hawking and n. turok , phys .
* b425 * , 25 ( 1998 ) .
a. vilenkin , phys . rev . *
d57 * , 7069 ( 1998 ) .
s.w . hawking and i.g .
moss , phys .
110b * , 35 ( 1982 ) . g. w. gibbons , s.w . hawking and m.j .
perry , nucl .
* b138 * ( 1978 ) , 141 . | the so - called negative mode problem in the path integral approach to the false vacuum decay with the effect of gravity has been an unsolved problem .
several years ago , we proposed a conjecture which is to be proved in order to give a consistent solution to the negative mode problem .
we called it the `` no - negative mode conjecture '' . in the present paper
, we give a proof of this conjecture for rather general models .
recently , we also proposed the `` no - supercritical supercurvature mode conjecture '' that claims the absence of supercritical supercurvature modes in the one - bubble open inflation model . in the same paper , we clarified the equivalence between the `` no - negative mode conjecture '' and the `` no - supercritical supercurvature mode conjecture '' .
hence , the latter is also proved at the same time when the former is proved . *
the no - negative mode theorem * * in false vacuum decay with gravity * takahiro tanaka _ department of earth and space science , graduate school of science _ _ osaka university , toyonaka 560 - 0043 , japan _ |
let @xmath9 be a lvy subordinator , that is , a non - decreasing lvy process starting from @xmath10 , which is continuous from the right with left limits .
this process has stationary and independent increments and is characterized by its laplace transform @xmath11 the function @xmath7 above is known as the lvy exponent ( or laplace exponent ) and is given by the lvy - khintchine formula : @xmath12 where @xmath13 is the drift and the lvy measure @xmath14 is a measure on @xmath15 which satisfies @xmath16 ( see @xcite , @xcite or @xcite ) .
consider the _
inverse subordinator _
@xmath17 , which is given by the first - passage time of @xmath8 : @xmath18 the process @xmath19 is non - decreasing , and its sample paths are a.s .
continuous if and only if @xmath8 is strictly increasing . also , @xmath4 is , in general , non - markovian with non - stationary and non - independent increments .
inverse subordinators are of particular interest in the study of fractional kinetics and the scaling limits of continuous time random walks , @xcite , @xcite , @xcite , @xcite , @xcite . here , a markov process , @xmath20 , is time - changed with an inverse subordinator , yielding the new process @xmath21 , @xmath22 . for certain choices of subordinators
@xmath8 , using @xmath4 as a time change gives a model of _ anomalous diffusion _ , where the mean - squared displacement of @xmath23 grows non - linearly in time .
for example , if @xmath8 is an @xmath24-stable subordinator , that is , the subordinator whose lvy exponent is given by @xmath25 then the mean of @xmath19 is given by the power law @xmath26 .
other types of non - linear behavior are also possible . in @xcite , meerschaert considers the subordinator with lvy exponent given by the following generalization of ( [ e : astable ] ) : @xmath27 where @xmath28 is a probability measure on @xmath29 . depending on the choice of @xmath28 ,
the mean of @xmath4 in this case can grow at logarithmic rates .
meerschaert also shows in this case that the transition density of the process @xmath23 solves a time - fractional diffusion equation whose order is distributed according to @xmath28 .
inverse subordinators have also appeared in many other areas of probability theory .
early work regarding the joint distribution of @xmath19 and @xmath30 was done in @xcite and @xcite .
more recently , in @xcite , kaj and martin - lf show that a scaled sum of these processes converges weakly to another non - stable and non - gaussian process .
an application of inverse subordinators in modeling of foreign exchange markets was considered in @xcite .
distributional properties of inverse subordinators were studied in @xcite , which drew upon a connection with cox processes . a different approach using differential equations to characterize the joint distribution function of the process @xmath4 was used in @xcite and @xcite .
an important function in the study of inverse subordinators is the so - called _ renewal function _ , @xmath31 , @xmath32 , which is given by the mean of the inverse subordinator , @xmath33 .
it has be shown by different methods that this function characterizes all finite - dimensional distributions of the process @xmath4 .
the laplace transform of @xmath5 is given simply in terms of the lvy exponent @xmath7 , however , inverting this laplace transform to obtain @xmath5 in closed form is not always possible .
in this paper , we provide two numerical methods for obtaining this mean first - passage time which take as input the drift @xmath34 and the lvy measure @xmath14 of the corresponding subordinator @xmath8 .
once @xmath5 is obtained , higher order moments can be calculated by methods described in @xcite or @xcite .
this paper is organized as follows : we begin in section [ s : background ] by giving a brief background on the theory of inverse subordinators , as well as describe the importance of the function @xmath5 . in section [ s : numerical ] , we develop two methods for calculating @xmath5 numerically and test them in cases where @xmath5 can be computed analytically . in section [ s : examples ] we apply these techniques to the following examples : poisson process with drift ( section [ s : poisson ] ) , compound poisson process with pareto jumps ( section [ s : pareto ] ) , special cases of the `` mixed '' @xmath24-stable process ( section [ s : astable ] ) , and the generalized inverse gaussian lvy process ( section [ s : gig ] ) .
for each example , we compute one and two - time moments of @xmath4 .
methods to compute @xmath5 for each example are implemented in matlab .
the software for doing this is freely available from the authors and its use is described in section [ s : guide ] .
we start by recalling the fundamental relationship between a subordinator and its inverse .
if @xmath8 is strictly increasing , then for @xmath35 and @xmath36 positive , @xmath37 if @xmath8 is not strictly increasing , then the above relationship holds off a set of measure @xmath10 in @xmath38 ( see @xcite ) . as in the introduction , define the _ renewal function _ to be the mean of the inverse subordinator , @xmath39 for @xmath32 .
also , let @xmath40 $ ] . from the lvy - khintchine formula , we have @xmath41 .
this together with ( [ e : fundeqn ] ) lets us compute the laplace transform , @xmath42 , of @xmath5 : @xmath43 e^{-\lambda t } ds dt \\ &
= & \int_0^\infty \int_0^\infty p[d(s ) < t ] e^{-\lambda t } ds dt \\ & = & \int_0^\infty \int_0^\infty \frac{1}{\lambda } e^{-\lambda t } dh_s(t ) ds \\ & = & \frac{1}{\lambda } \int_0^\infty e^{-s \phi(\lambda ) } ds = \frac{1}{\lambda \phi(\lambda)}. \label{e : lt_mfpt}\end{aligned}\ ] ] thus , @xmath42 characterizes the process @xmath4 ( since @xmath7 characterizes @xmath8 ) .
since @xmath4 is non - decreasing , we can define the borel measure @xmath44 which is induced by @xmath5 , which is commonly referred to as the _ renewal measure_. notice that the renewal measure has the following property . for a.e .
@xmath45 , we have @xmath46}(\tau ) du(\tau ) & = & u(b ) - u(a ) \nonumber \\ & = & \int_0^\infty p[e(b ) > s ] - p[e(a ) > s ] ds \nonumber \\ & = & \int_0^\infty p[d(s ) \leq b ] - p[d(s ) \leq a ] ds \nonumber \\ & = & \mathbb{e } \int_0^\infty \left ( { \mbox{\boldmath$1$}}_{(-\infty , b]}(d(s ) ) - { \mbox{\boldmath$1$}}_{(-\infty , a]}(d(s ) ) \right ) ds \nonumber \\ & = & \mathbb{e } \int_0^\infty { \mbox{\boldmath$1$}}_{(a , b]}(d(s ) ) ds .
\label{e : renewalmeasure}\end{aligned}\ ] ] by approximating with step functions , this relationship can be extended to continuous functions @xmath47 as @xmath48 . with this
, we see the laplace transform of @xmath44 is given by @xmath49 the pair @xmath5 and @xmath44 can be used to compute all joint moments of the process @xmath4 . for non - negative integers @xmath50
define @xmath51 the order of the moment in ( [ e : nmoment ] ) is defined to be @xmath52 .
the following theorem from @xcite gives the @xmath53-time laplace transform @xmath54 in terms of a strictly lower older moment .
[ t : inversemoments ] let @xmath8 be a general lvy subordinator with lvy exponent @xmath7 and let @xmath4 be the inverse subordinator of @xmath8 .
the @xmath53-time laplace transform of the @xmath55 order moment @xmath56 defined in ( [ e : nmoment ] ) is given by @xmath57 notice that the laplace transform of @xmath56 is given as the product of @xmath58 and the laplace transform of a strictly lower order moment .
taking inverse laplace transforms , the @xmath55 order moment is given as the sum of convolutions @xmath59 for full details , see @xcite .
thus , if one knows the function @xmath60 , @xmath32 , then all higher order moments can be obtained inductively .
for example , the covariance is given by @xmath61 it is not always easy , however , to invert the laplace transform ( [ e : lt_mfpt ] ) to obtain @xmath31 analytically . in the following ,
we give two methods for numerically inverting this laplace transform given only the drift @xmath34 and lvy measure @xmath14 of the subordinator @xmath8 .
once @xmath5 is obtained , the density of the renewal measure can be obtained by numerical differentiation and integral expressions such as ( [ e : inversecov1 ] ) can be approximated by numerical integration .
the first method is based on the bromwhich integral which expresses the inverse laplace transform as a path integral in the complex plane , and the second is based on the post - widder inversion formula , which expresses @xmath5 as a limit of terms involving derivatives of @xmath42 .
in section [ s : background ] we saw that all moments of an inverse subordinator can be computed if one has first computed the renewal function @xmath39 , which is given by the inverse laplace transform of ( [ e : lt_mfpt ] ) . in some cases , an analytical expression for @xmath5 can be found , and in most cases , the asymptotics of @xmath5 can be studied using a tauberian theorem . in this section
we give two methods for calculating @xmath5 numerically .
the first is simple and precise , but is difficult to use when @xmath7 is a complicated function .
the second is more robust , but requires smoothness in @xmath5 to be effective .
the first method involves calculating the inverse laplace transform of @xmath42 by numerically approximating the bromwich integral : @xmath62 where @xmath63 is chosen such that @xmath42 is analytic in the region @xmath64 .
using the fact that @xmath65 for @xmath66 , and symmetry properties of analytic functions which are real valued on the real axis , ( [ e : bromwich ] ) simplifies into two equivalent expressions : @xmath67 for details see @xcite .
since @xmath70 and @xmath7 is increasing in @xmath71 , @xmath6 has a singularity the origin .
therefore , we much choose @xmath72 above . then , given @xmath34 and @xmath14 , we evaluate the integrals in ( [ e : phirphii ] ) to obtain @xmath73 and @xmath74 and use ( [ e : realut ] ) and ( [ e : imut ] ) to obtain @xmath75 and @xmath76 for fixed @xmath77 and then compute either integral in ( [ e : integral ] ) to get @xmath31 . alternatively ,
if @xmath7 is known in closed form , then @xmath73 and @xmath74 can be computed directly .
this method works fairly well when @xmath73 and @xmath74 are easy to compute , for instance , in the case of the poisson process with drift .
the main problem with the method is when the integrands in ( [ e : integral ] ) and/or ( [ e : phirphii ] ) are highly oscillatory and slowly decaying , causing most integration algorithms to converge very slowly .
the following method is based on the post - widder inversion formula ( @xcite , theorem 2 or @xcite , section vii.6 ) . in order to justify using this formula in our case , we state and prove this result under slightly weaker conditions than those found in @xcite .
( post - widder inversion ) .
let @xmath78 be a continuous function such that @xmath79 as @xmath80 for some @xmath81 , and let @xmath82 be the laplace transform of @xmath77 . then for @xmath83 , we have @xmath84 where @xmath85 denotes the @xmath86 derivative , @xmath87 .
fix @xmath83 .
let @xmath88 , @xmath89 be iid gamma random variables with mean @xmath90 and variance @xmath91 and let @xmath92 .
the idea is to approximate @xmath93 by @xmath94 .
notice that @xmath95 has a gamma distribution with mean @xmath90 and variance @xmath96 . by the law of large numbers , @xmath97 as @xmath98 and by the continuous mapping theorem ( @xcite , theorem 5.10.4 ) , @xmath99 .
observe that we also have the following convergence in mean : @xmath100 which follows from the uniform integrability of @xmath101 .
this can be seen using the density of a gamma random variable and the assumption that @xmath102 as @xmath80 .
indeed , we have @xmath103 * 1 . * in order to justify using the post - widder formula , we must be sure that the renewal function @xmath5 satisfies @xmath105 as @xmath106 .
fortunately , the renewal theorem ( see for instance , proposition 3 in @xcite ) implies that if @xmath8 has finite mean , then @xmath31 will grow as @xmath107 as @xmath108 .
if @xmath8 has infinite mean , its mean first hitting time @xmath109 will , on average , grow slower than in the finite mean case for large @xmath90 . to see this , any subordinator @xmath8 with infinite mean
can be bounded below by a subordinator @xmath110 with finite mean by truncating the large jumps of @xmath8 for all @xmath111 a.s . ] .
the inverse , @xmath112 , of @xmath110 will be a.s .
greater than @xmath4 , and thus @xmath113 .
we will first focus on step @xmath119 above .
the main difficulty with using this method resides in the fact that ( [ e : pwinversion ] ) involves derivatives of arbitrarily high orders , which are often difficult to compute .
fortunately , in our case , this calculation can be done in a reasonable fashion , which we now outline .
recall leibnitz s formula , which states that for smooth functions @xmath120 , @xmath121 let @xmath122,@xmath83 , and @xmath123 . applying leibnitz s formula to @xmath124 with @xmath125
gives @xmath126 where the vectors @xmath127 are given by @xmath128 to compute the components of @xmath129 , we use an idea from @xcite . using leibnitz s formula on the unit function @xmath130
, we see that for any @xmath131 , @xmath132 from this , we obtain the following matrix equation : @xmath133 choosing @xmath134 , we conclude that @xmath129 satisfies the matrix equation @xmath135 , where @xmath136 and @xmath137 observe that the entries in @xmath138 are easily expressed in terms of the drift @xmath34 and the lvy measure @xmath14 corresponding to the subordinator @xmath8 .
indeed , we have @xmath139 thus , to compute @xmath129 , one needs to compute the entries of @xmath138 using ( [ e : gji ] ) and ( [ e : phidiff ] ) , and then solve the matrix equation @xmath135 .
notice that the integrals in ( [ e : phidiff ] ) are much easier to compute than those in section [ s : integration ] since the integrands do not oscillate ; they decay exponentially and are positive .
observe that @xmath140 , and hence @xmath141 are easy to compute . using this method , we get @xmath142 in ( [ e : uks ] ) .
the density @xmath143 of the renewal measure is also of interest , for example in ( [ e : inversecov1 ] ) .
it can be approximated as @xmath5 is in ( [ e : uks ] ) , since the laplace transform of @xmath143 is @xmath144 .
one needs to compute @xmath145 , which is obtained with no extra cost from ( [ e : matrixeqn ] ) since it is the last component of @xmath129 . for step 2 above , we refer to the technique explained in @xcite .
to summarize , it has been shown ( @xcite,@xcite ) that if @xmath5 is smooth , then we have the following series expansion : @xmath146 where @xmath147 are remainder terms .
write @xmath148 and let @xmath149 . with this , ( [ e : seriesexp ] ) implies @xmath150 our goal is to compute @xmath151 with @xmath90 fixed . to do so
, we consider @xmath152 as a function of @xmath153 and , given @xmath154 and @xmath155 , we write down the so - called `` lagrange polynomial '' @xmath156 : this is the polynomial of degree @xmath157 which passes through the @xmath53 points @xmath158 , @xmath159 , @xmath160 observe that if @xmath161 for some @xmath162 , then all the summands in ( [ e : lagrangepoly ] ) vanish except for the term where @xmath163 , which is equal to @xmath164 .
thus the polynomial @xmath156 passes through the @xmath53 points @xmath158 , @xmath159 .
observe that we can not compute @xmath165 since @xmath166 corresponds to @xmath167 .
however , with the method described above , we can compute @xmath168 for @xmath169 and then approximate @xmath170 by @xmath171 . since @xmath172 , @xmath5
is then approximated by the linear combination @xmath173 where the weights @xmath174 , calculated by setting @xmath166 in ( [ e : lagrangepoly ] ) , depend on the choice of @xmath175 ) obtained by polynomial interpolation is equivalent to taking the linear combination which cancels the first @xmath53 remainder terms in ( [ e : seriesexp ] ) . ] .
thus , given a desired level of accuracy @xmath181 and @xmath83 fixed , we compute the sequence @xmath182 until @xmath183 and then set @xmath184 . typically this method converges with @xmath185 for reasonable values of @xmath186 . since @xmath187 , it is often impossible to compute @xmath188 for @xmath189 in double precision arithmetic , as noted in the remarks below .
note that for ( [ e : seriesexp ] ) to hold and for this method to be most effective , @xmath5 must be sufficiently smooth .
we found that for points at which @xmath5 is not differentiable , the sequence @xmath190 had a slower rate of convergence .
caution should be taken when computing terms like @xmath192 for @xmath193 large , since numbers like @xmath194 and @xmath195 are outside the range of double precision arithmetic .
instead , one should use expressions like @xmath196 , because while @xmath197 and @xmath198 may be large , their ratio may be small .
nevertheless , for large enough @xmath177 and @xmath199 ( relation ( [ e : vk ] ) , for example , which requires computing @xmath200 ) , even the ratio may be too large .
this happens for instance if @xmath201 , in which case @xmath202 . to be safe , we used at most @xmath187 .
a similar overflow problem can occur for @xmath203 large . to avoid this
, one can make the following adjustment to ( [ e : pwinversion ] ) .
let @xmath204 , then @xmath205 where @xmath206 is a rescaled version of @xmath42 : @xmath207 repeating the steps of this method with ( [ e : pwinversion2 ] ) , we get @xmath208 where @xmath209 and @xmath210 is the solution to the matrix equation @xmath211 , where @xmath212 . to compute these products and ratios , a procedure as the one describes in the previous remark should be done to avoid overflow .
we found the choice @xmath213 useful . in table
[ t : errors ] , the post - widder method is tested in cases where @xmath5 can be computed explicitly ( see section [ s : examples ] for explanations of each example ) .
the values in the table were obtained using a threshold of @xmath214 . with the exception of starred entry in table [ t : errors ] , all converged within this threshold with @xmath215 .
if sufficient convergence did not occur , @xmath216 is used as the approximation . in these cases ,
the approximation is very good . for the poisson process
, @xmath5 is discontinuous at the integers .
therefore , we used instead @xmath217 . since the post - widder method does not apply when @xmath5 is discontinuous
, we do not expect to get good results in the poisson case . and indeed , the absolute errors are large .
for instance , table [ t : errorsint ] shows that when @xmath218 , the algorithm converged but to the wrong value ( the absolute error between the true and computed value is @xmath219 ) .
the plot in figure [ f : pwcompare ] shows in more detail the erratic behavior of the post - widder method in this case .
notice that for larger times this method approximates @xmath5 ( which in this case is a step function ) with a straight line .
.absolute errors between exact values of @xmath5 and those given by this method for a selection of subordinators whose mean first passage time can be computed exactly .
here we used a threshold of @xmath220 . [ cols="<,^,^,^,^,^",options="header " , ] j. abate , g.
l. choudhury , and w. whitt .
an introduction to numerical transform inversion and its application to probability models . in w.
k. grassman , editor , _ computational probability _ , pages 258322 .
kuwer academic publishers , usa , 2000 .
ernst eberlein and ernst august v. hammerstein .
generalized hyperbolic and inverse gaussian distributions : limiting cases and approximation of processes . in _
seminar on stochastic analysis , random fields and applications iv _ , volume 58 of _ progr .
_ , pages 221264 .
birkhuser , basel , 2004 . | let @xmath0 be a non - decreasing lvy process .
the first - hitting time process @xmath1 ( which is sometimes referred to as an inverse subordinator ) defined by @xmath2 is a process which has arisen in many applications .
of particular interest is the mean first - hitting time @xmath3 .
this function characterizes all finite - dimensional distributions of the process @xmath4 .
the function @xmath5 can be calculated by inverting the laplace transform of the function @xmath6 , where @xmath7 is the lvy exponent of the subordinator @xmath8 . in this paper , we give two methods for computing numerically the inverse of this laplace transform .
the first is based on the bromwich integral and the second is based on the post - widder inversion formula .
the software written to support this work is available from the authors and we illustrate its use at the end of the paper . |
the observation of cosmic rays with energy higher than the greisen - zatsepin - kuzmin ( gzk ) cutoff at @xmath3 ev @xcite implies a corresponding flux of ultra high - energy ( uhe ) neutrinos , with energy in the @xmath4 ev range @xcite .
these secondary neutrinos are created via photomeson production of cosmic rays on the 2.7 k cosmic microwave background .
detection of these neutrinos would provide unique information about the origin of primary cosmic rays and the nature of their sources .
although predictions for the flux of uhe neutrinos differ by orders of magnitude , a reasonable set of parameters puts the rate of uhe neutrinos at the level of 10/km@xmath5/century @xcite .
a detector volume of hundreds of cubic kilometers water equivalent is required to detect a significant number of neutrinos in this energy regime .
optical techniques are widely used in neutrino detectors , but the volume that can be instrumented is constrained by attenuation lengths of tens of meters at optical frequencies in detector media ( such as ice ) , so optical detectors have limited sensitivity to rare high - energy events .
g. askaryan first predicted coherent radio emission from ultra high - energy showers @xcite , and the effect has been confirmed in accelerator beam tests @xcite .
the energy of the coherent radio emission that results from the development of a negative charge excess in the shower depends quadratically on shower energy , and for showers with energy greater than @xmath6 ev , dominates the emitted cherenkov power spectrum .
askaryan also proposed a few materials as detection media that occur naturally in very large volumes and are expected to have long attenuation lengths in the radio regime , including ice and salt .
radio attenuation lengths longer than 1 km have been measured in ice near the south pole @xcite .
formations of salt rock could therefore be a viable detector for high - energy neutrinos if attenuation in the radio regime is indeed low .
domes of rock salt occur naturally in many parts of the world , including the gulf coast region of the united states . in these formations ,
the salt originates from dried ocean beds which have been buoyed upward due to geological forces through a process called diapirism . through this process
, the salt becomes very pure as impurities are extruded .
such salt domes , with typical dimensions of several square kilometers by several kilometers deep , are thought to be good candidates for a neutrino detector . for a more detailed discussion of the dielectric properties of salt and the application to neutrino detection , see reference @xcite .
there are two classifications of measurements that have been made of the radio properties of rock salt in salt domes .
ground penetrating radar ( gpr ) was used in the 1960 s and 1970 s to determine the size and structure of salt domes .
this technique is based on sending a radar signal into the salt and looking for reflections from interfaces in the salt structure .
one can calculate the distance of the interface from the time delay of the reflected pulse . although the primary purpose of the gpr measurements is to calculate interface distances
, one can also extract an attenuation coefficient based on the detected signal voltage , the voltage of the transmitted pulse , and the distance over which the pulse traveled . using the gpr technique to extract attenuation
is complicated by the unknown reflection coefficient and geometry at the surface of reflection , but assuming a coefficient of @xmath7 from a flat surface gives a conservative estimate .
direct measurements of attenuation in rock salt have also been made . in 2002 ,
measurements made at hockley salt mine showed attenuation lengths consistent with being longer than 40 m at 150 and 300 mhz @xcite .
their measurements were limited by the voltage of the pulser , which limited their transmission distances in the salt .
we decided to follow the techniques of the direct transmission measurements but used a high voltage pulser so that we could transmit through longer distances of salt .
we made measurements of the dielectric properties of salt in the mine located in the cote blanche salt dome in st .
mary parish , louisiana in august 2007 .
we chose this dome because gpr measurements that were made in the mine suggested very low radio attenuation , including observations of reflections over the longest distance of any mine measured @xcite .
the cote blanche dome is one of five salt domes in the area .
the salt dome extends from approximately 90 m below the surface to 4270 m below the surface . at a depth of 1100 ft .
( 335 m ) , the salt extends 1700 m east to west and 2100 m north to south . at 2000
( 610 m ) deep , the horizontal cross section has nearly doubled in size .
the total volume of salt in the dome is estimated to be 28 - 30 km@xmath8 @xcite .
the dome has been actively mined since 1965 using the conventional `` room and pillar '' method .
the mine consists of a grid of 30 ft .
( 9 m ) wide by 30 ft .
( 9 m ) high drifts ( corridors ) spaced by 100 ft . by 100 ft .
( 30 m by 30 m ) pillars of salt , and has three levels at depths of 1100 ft .
, 1300 ft . , and 1500 ft .
( 335 m , 396 m , and 457 m ) .
each level covers several square kilometers .
current mining operations are on the 1500 ft .
deep level .
the measurements described in this paper are from the third trip that we made to the cote blanche mine . in may 2005 , on our first visit to the site , with m. cherry and j. marsh from louisiana state university
, we established the viability of the experimental setup and measured field attenuation lengths along a corridor ( within 10 meters of a wall ) on the 1300 ft .
level of the mine , of approximately 10 - 15 m at 150 mhz and 300 mhz @xcite . in september 2006 , on a return visit to the mine , we transmitted and received signals horizontally across a single pillar of salt , both with antennas placed against the walls and with antennas inserted 4 m into the ceiling in shallow boreholes .
we measured attenuation lengths of @xmath9 m in the frequency range 50 - 150 mhz , @xmath10 m at 150 - 250 mhz , and @xmath11 m at 250 - 350 mhz .
we also measured an average index of refraction of @xmath12 @xcite .
the attenuation lengths that we measured were inconsistent with previous gpr studies of the cote blanche mine ( see section [ sec : gpr ] ) .
the relatively short attenuation lengths that we measured through the walls of the pillar could be due to the method used to mine the salt .
the miners carve out corridors in the salt by cutting a horizontal slice out from under the wall and then blasting the section above the floor so that the salt can be removed , and then proceed to a new section of salt further along the corridor . the data from our first two trips were consistent with a model of very lossy salt ( approximately 10 m attenuation length ) in the region closest to the walls of the pillar and longer attenuation lengths ( quoted above ) in salt more than about 10 m from the wall .
figure [ fig : saltmap ] shows the region of the 1500 ft . deep level of the mine where we made our measurements , and figure [ fig : systemdiagram ] is a diagram of the experimental setup .
the miners drilled three 2.75 inch ( 7.0 cm ) diameter boreholes into the salt of the floor of the 1500 ft .
level of the mine .
boreholes 1 and 2 were 100 ft .
( 30 m ) deep , while borehole 3 was 200 ft .
( 60 m ) deep . at the first two boreholes ,
drilling was stopped at 100 ft . once the driller encountered methane gas .
we used a fast , high - power , broadband pockel s cell pulser model hyps from grand applied physics with a peak voltage of 2500 kv and a 10%-90% rise time of 200 ps to generate radio pulses .
there was 5 ft . of lmr 240 cable and 200 ft . of lmr 600
cable leading to the transmitting as well as from the receiving antennas , and the antennas were lowered by hand into the salt .
we made most of our measurements using three pairs of dipole antennas .
we measured the 3 db points of each antenna when they were in a borehole in the salt to determine the in - band frequencies .
the 3 db points of the transmission of the low frequency ( lf ) antenna pair ( raven research rr6335 ) , are 50 to 175 mhz in salt . for the medium frequency ( mf ) antennas , which we custom made ,
the 3 db points are 175 and 500 mhz in salt .
the high frequency ( hf ) pair ( shure incorporated ua820a ) has a transmission band between 550 and 900 mhz in salt .
we used a tektronix tds694c oscilloscope to record the received signal pulse .
we used a pulser with synchronized outputs to trigger both the the high - power pulser and the oscilloscope .
this allowed us to look in a fixed time window for the signal and reduce noise by averaging many waveforms .
the transmitting antenna was always lowered into borehole 1 , and the identical receiving antenna was either in borehole 2 , which was 50 m away from the transmitter , or borehole 3 , which was 169 m away from the transmitter .
we took data at 10 ft .
( 3 m ) incremental depths in the boreholes .
the setup used to measure transmission between boreholes 1 and 2 was identical to the setup used to measure between boreholes 1 and 3 so that we could make reliable relative measurements without requiring an absolute system calibration . because the transmission band of each antenna was relatively broad
, we were able to make measurements between 125 and 900 mhz in salt .
figure [ fig : timedomain12_13 ] shows an example waveform of the received pulse at borehole 2 and borehole 3 using the mf antennas , and figure [ fig : freqdomain12_13 ] shows the fourier transform of the same waveforms . to calculate the attenuation length , we first cut the recorded waveforms in a time window around the pulse to eliminate any reflections and reduce contributions from noise .
the width of the window is 35 ns for the lf antennas , 30 ns for the mf antennas , and 12 ns for the hf antennas .
we expect to record reflections from interfaces within the salt and from the salt - air boundary .
we were able to see reflections from the 1500 ft .
corridor with the receiver at either borehole .
the time window that we use to for the analysis extends no more than 20 ns after the peak of the pulse to ensure that we eliminate the possibility of interference from any reflection off of the corridor for depths of 50 ft .
and deeper .
we sum the total power in frequency bands that are 50 mhz wide for the lf and mf antennas , and 100 mhz wide for the hf antennas which had a larger bandwidth .
we subtract the noise contributions in each band .
the noise power is taken from a sample waveform for each antenna type in a time window of the same length as the pulse window but earlier in time than any pulse appears .
the noise subtraction had a small effect on the calculated attenuation length because the signal to noise ratio was high in the band of the antenna .
we define voltages @xmath13 and @xmath14 to be the square root of the power in the @xmath15 frequency bin between boreholes 1 and 2 and boreholes 1 and 3 which span distances @xmath16 and @xmath17 respectively .
since we used the same system regardless of the location of the receiver , the voltages received in the boreholes from the transmitter are related by : @xmath18\ ] ] where @xmath19 is the field attenuation in the @xmath15 frequency bin .
inverting this equation gives an expression for the field attenuation length in each frequency bin : @xmath20
the main source of systematic uncertainty on the measurement of the field attenuation length at a given depth is due to the position of the antenna within the hole . in previous trips to cote blanche , we discovered that a small variation in position of the antenna led to a significant change in the voltage received through the salt @xcite .
we estimate the size of this uncertainty as the root mean square variation between neighboring depths of the peak - to - peak voltage of the recorded waveform at each depth ( excluding the 10 and 20 ft . depths ) : @xmath21 where @xmath22 is the number of depths included here . using this method
, we estimate the uncertainty on the voltage measured due to the position of the antenna in the hole to be @xmath23 .
we also include an uncertainty due to the exact choice of the time window that contains the pulse .
we estimate this uncertainty as the root mean square variation of the total power in the pulse as we slide the time window by 1 ns . when the power at a given frequency is small , this uncertainty dominates ( up to 50% in voltage ) , but in the frequency band of the antenna , the uncertainty is small ( less than 10% in voltage ) . there is also an uncertainty on the distance between the holes .
we measured the distance between boreholes 1 and 2 with a measuring tape , and then used relative timing of received pulses to extrapolate the distance between boreholes 1 and 3 .
the uncertainty on these distances include a contribution from system timing ( @xmath24 ns ) , one due to the measurement of the distance between boreholes 1 and 2 ( @xmath25 ft . ) , and one due to a depth - dependent potential deviation of the boreholes from vertical ( @xmath26 ft . ) .
the maximum uncertainty on the distance between the transmitting and receiving antennas is always less than @xmath27 .
using the same pulser that was used for the attenuation measurements , we measured the fraction of power reflected from each antenna while it was in a borehole so that we could deduce the fraction transmitted and frequency dependence . for this s11 measurement
, we inserted a coupler ( minicircuits zfdc-20 - 4 ) between the pulser and antenna and recorded the reflected signal through the coupled port .
after measuring the reflection from the open cable ( with the setup identical but with only the antenna removed ) , the ratio of the power in the two pulses is the fraction reflected from the antenna .
the transmitted fraction is deduced by taking the sum of the reflected and transmitted power fractions to be unity .
figure [ fig : s21 ] shows the transmission of an mf antenna as a function of frequency .
the transmission did not change significantly when we changed the depth of the antenna in the hole .
figure [ fig : depth ] shows the attenuation length in the frequency bin centered on 250 mhz as measured with the mf antennas plotted versus depth .
we took measurements at 10 and 20 ft .
depths , and the values we calculate for the attenuation length at those shallow depths are consistent with attenuation lengths at greater depths ; however , we only report the results from 30 ft . and deeper since the deeper measurements are more likely to be of unfractured , clear salt .
at 30 and 40 ft .
, as much as 1/4 of the power in the waveforms received at borehole 3 may be due to power from a reflected pulse ( based on measurements at greater depths ) but we still include those points here .
we do not observe a depth dependence in attenuation length .
measured field attenuation length at 50 ft . and 90 ft .
depths are shown in figure [ fig : freq ] as a function of frequency .
the lf antennas were not easily lowered beyond 75 ft .
depth , so we include data from those antennas at 75 ft . instead of 90 ft .
if the salt had a constant loss tangent , we would expect that the attenuation length would decrease with increasing frequency as @xmath28 , where @xmath29 is the frequency of the radiation @xcite .
we fit the points in figure [ fig : freq ] to a power law of the form : @xmath30 and the best fit values are @xmath31 m and @xmath32 , with a @xmath33/dof=25.6/36 .
although it is clear that the attenuation length is falling with increasing frequency , it does not follow the @xmath28 expectation , which hints at a non - constant loss tangent , consistent with measurements at the hockley mine @xcite . in table [
tab : attnlengths ] , we list the measured attenuation lengths at a few frequencies of interest based on the best fit values from this fit .
the uncertainties on the distances between the holes is one that would move the measured attenuation length in the same direction at all measurement positions , so we do not include them in the fit .
the uncertainties on the distances between the holes are small compared to the remaining uncertainties .
if we vary the distances by their uncertainties and refit , the attenuation lengths change at most by @xmath34% .
we have also averaged the attenuation length at each frequency over all depths of 50 ft .
and deeper and over all antenna types , and plotted the results in figure [ fig : avg ] .
this figure also shows the best fit power law function , with best fit values @xmath35 m and @xmath36 with a @xmath33/dof=13.3/12 . for this plot
the uncertainties only include the scatter between measured attenuation lengths using different antennas and at different depths in the same frequency bin .
the attenuation lengths that we measured on this trip are significantly longer than those that we measured on our previous trip to the same mine .
the measurements that we report here were made in boreholes that were drilled into the floor of the lowest level of the mine , whereas the previous measurements were made with holes drilled into the ceiling and against the walls of the corridors . because the method of mining consists of cutting under the salt and then blasting the wall above the cut , any fractures that occur due to the process would tend to propagate upward .
this means that the salt in the floor of the lowest level of the mine would tend to be less fractured and probably exhibit less loss in the radio regime due to scattering .
we also calculate the index of refraction of the salt using the direct transmission measurements that we made .
the index of refraction is defined as @xmath37 , where @xmath38 is the real part of the dielectric permittivity .
we calculate the speed of transmission through the medium using the distance between boreholes 1 and 2 and the time of travel of the signal pulse through the salt .
the time of travel was measured by taking the difference between the time that the pulse was generated and the signal was received and subtracting the known system delay .
figure [ fig : n ] shows the measured index of refraction at each depth .
the index of refraction that we measure is consistent with @xmath39 , the index of refraction of rock salt .
we estimate an uncertainty of @xmath24 ns on the absolute system timing and @xmath25 ft . on the distance between the holes .
we also include the same uncertainty due to potential deviation of the boreholes from vertical that was described in section [ sec : error ] , which contributed @xmath26 ft . to the uncertainty . at nearly every measurement position we observed at least one
clear `` reflected '' signal in addition to the `` direct '' signal that traveled along the straight line between the transmitting and receiving antennas . reflected
signals can be easily distinguished from direct signals from their time of arrival at the receiver .
they allow us to probe distances greater than that between our drilled holes , and to probe different salt regions as well .
however , little is known about the loss in power incurred at the reflection , and reflected signals are often transmitted and received by the antennas at oblique angles , where the antenna response is less well understood . with our antennas in boreholes 1 and 2 ,
the entirety of the shortest path between the receiver and transmitter was below the corridor where we were working , so we expected to see reflections from that corridor .
we did observe signals that were consistent with this interpretation .
the measured time differences between the direct and reflected signals while we were transmitting between boreholes 1 and 2 were within approximately 10 ns of the expected time difference at all depths .
these reflected pulses traversed as much as 256 ft .
( 78 m ) of salt , and a discrepancy of 10 ns corresponds to approximately 4 ft .
( 1.2 m ) in salt . with the antennas at the greatest depths , most of the power in these reflected signal were out of the band of the antenna .
we conclude that this power was reflected off of the input of the antenna and radiated from the feed cable which was acting as a long - wire antenna .
in addition , the expected beam pattern of a long - wire antenna is consistent with the angle of emission seen with these reflections . with our antennas in boreholes 1 and 3
, the specular path of a reflection from the corridor level is not incident on a salt - air boundary , but nevertheless at depths greater than 50 ft .
( where the reflections did not interfere with the direct signal ) we did observe signals whose timing were consistent with reflections from the corridor level within 10 ns .
in addition , with the antennas in boreholes 1 and 3 we observed signals at all depths 30 ft . and
deeper consistent with having reflected from 170 ft . above the corridor where we stood .
the 1300 ft .
level of the mine is of course nominally 200 ft . above the floor of the corridor .
therefore , we observed signals having traveled through as much as 624 ft . of salt when the mf antennas were 90 ft . and 150 ft .
deep .
the deep holes provided to us by the cote blanche mine presented a unique opportunity to measure the beam pattern of our antennas while completely submerged in salt , although due to the distances involved , we did not probe a broad range of angles . with one mf antenna at 90 ft .
depth in borehole 1 , we took data with the second mf antenna in borehole 2 and at depths of 20 ft . , 50 ft . , 70 ft . and 90 ft . , and in borehole 3 at depths of 90 ft .
, 130 ft . and 150 ft .
the waveform at 20 ft .
depth contained interference from a reflected signal , so that data was not included in this measurement .
the remaining data probed the antenna beam pattern at angles of -6.2@xmath40 , -4.1@xmath40 , 0@xmath40 , 7.0@xmath40 and 13.8@xmath40 with respect to the horizontal , defined to be positive when the antenna in borehole 1 was deeper than the antenna in the other borehole . after correcting for @xmath41 power loss as well as salt attenuation loss based on our measured values , we plot the measured power at each position relative to that measured at 0@xmath40 in the same borehole .
we use the same uncertainties as described in section [ sec : error ] .
the solid line in the figure is the expected beam pattern for a half - wave dipole in air : @xmath42^{2}.\ ] ] although our uncertainties are large , we do not see a deviation from the half - wave dipole beam pattern .
the data is also not inconsistent with the peak of the transmission being in the horizontal direction .
ground penetrating radar ( gpr ) measurements were made in the cote blanche mine and other salt mines in the 1970 s by stewart and unterberger @xcite to probe discontinuities in the salt and the location of the top of the dome .
they used a single frequency waveform generator at 440 mhz , a pair of high - gain antennas that were pointed into the salt of interest , and an oscilloscope to measure the time delay between the transmitted signal and any received reflections off of discontinuities deep within salt . to calculate the distance of the discontinuities
, they used an index of refraction measured via transmission across a pillar of salt ( their result implies @xmath43 ) .
they measured the location of the top of the cote blanche dome from within the mine by transmitting the signal vertically through the ceiling . although the paper reports that attenuation of 2 - 3 db per 100 ft .
is typical for radar in dry salt in general , which would imply field attenuation lengths in the range 90 - 140 m , at one measurement station they report a multiply - reflected signal that had traveled through a total path length of 4080 ft .
( 1244 m ) and stated that it was the longest transmission distance observed in any mine .
we have compiled information in table [ tab : gprspecs ] from the discussion of the specifications of the system used to make the measurements in references @xcite and @xcite .
the power received , @xmath44 , is related to the power transmitted , @xmath45 , by the friis formula : @xmath46 where @xmath47 and @xmath48 are the gain of the transmitting and receiving antennas , @xmath49 is the wavelength of the transmitted signal in salt , and @xmath50 is the distance between the antennas . using the result of the friis formula together with the gpr system specifications
, we estimate that the minimum attenuation length allowed to detect the reflected signal over the longest distance observed ( see table [ tab : gprspecs ] ) is 138 m , assuming that they detected the signal just at the sensitivity threshold of their system .
this attenuation length is not inconsistent with some of our measurements in the same frequency range , although our results show generally higher losses .
note that if there are variations in the clarity of the salt , the longest observed path length quoted in their paper would be more likely to be from the clearest salt that they had sampled among all of their measurements .
we only sampled one section of salt with these measurements . in order to replicate the low - loss results of the gpr technique
, any further measurements would have to be made with a portable high - power system capable of making measurements at many locations .
we made a brief attempt at making measurements using the gpr technique during our visit to the mine .
using the same high - power pulser and oscilloscope , we used a pair of directional antennas in an attempt to see reflections off of interfaces within the walls of the mine .
we did see late reflected pulses , but we did not understand that data well enough to conclude whether the reflections that we saw came from within the salt , or were merely reflections down the corridor of the mine .
we have run a simulation to estimate the sensitivity of an array of antennas embedded in salt with the attenuation lengths that we have measured here . as much as possible
, we use the same parameters as the simulated array described in the hockley paper for direct comparison @xcite .
the simulated array consists of 10@xmath5110@xmath5110 dipole antennas with center frequency at 150 mhz and 50% bandwidth in a salt formation that is a cube 4 km on a side .
we require 4 antennas hit with a signal - to - noise ratio of 4@xmath52 .
the previous study in @xcite considered a 300 m attenuation length at 300 mhz and a @xmath53 dependence , whereas we have measured a @xmath54 m attenuation length at 300 mhz and the spectral index that we measure is @xmath55 . reproducing the simulated hockley array
, we expect 12 events per year compared to the order 10 events per year that they quote in their paper . using the measured cote blanche attenuation lengths
, we expect 1.4 events per year ultra - high energy neutrino flux @xcite , for a factor of 9 reduction in sensitivity due to a lower measured attenuation length , but still a non - negligible rate over a one year run time .
we have measured field attenuation lengths in the frequency range 125 - 900 mhz using broadband pulses transmitted and received by dipole antennas in boreholes drilled 100 ft . into the floor of the 1500 ft .
deep level of the the cote blanche salt mine .
the best fit power law that describes the measured frequency - dependent attenuation lengths is given by @xmath56 with statistically dominated uncertainties .
this is the most precise measurement of radio attenuation in a natural salt formation to date .
the main result of this paper is derived from data that we took on our third trip to the cote blanche salt mine .
the data taken from the three trips indicates that the room and pillar system of mining leads to regions of salt near the walls and ceilings of the corridors with short attenuation lengths ( @xmath57 m ) which could be due to fracturing brought about by the explosives used in that method .
we find that the salt that is just under the floor of a corridor does not suffer the same degradation , which may be due to the cut along the floor that is made prior to the detonation of the explosives to blast away the salt .
although we expect there to be variation in salt properties between salt domes as well as regions of salt within the same dome , we have compared our measurements with previous measurements made at cote blanche and elsewhere .
our result is not inconsistent with attenuation lengths measured at the hockley salt mine .
we have attempted to compare our measurements with the longest transmission distance reported by gpr experts in the same mine .
our data generally showed greater attenuation than can be reconciled with that result .
however , due to the variation observed in our data , some of our measurements do point to salt that could have been clear enough to be consistent .
that we strain to reconcile our numbers with the longest transmission distance reported there may indicate that there are regions of the cote blanche mine with salt that has less attenuation than in the regions where our measurements were made . we have modeled a salsa array of @xmath58 dipole antennas in boreholes that reach 3 km depth to assess the sensitivity of a neutrino detector in salt with the attenuation lengths that we report here to compare with the simulation study described in the hockley attenuation length paper .
we find the lower best fit attenuation lengths measured at cote blanche reduce the sensitivity of such a detector measured in terms of event rate by a factor of 9 , but still maintain an expectation of order one event per year using a standard model for the expected ultra - high energy neutrino flux . in order for a next - generation neutrino detector in salt that measures @xmath57 gzk events / year to be feasible , attenuation lengths longer than the @xmath59 m lengths reported here
will likely be necessary .
however , since this result is based on a limited region of the dome , and past gpr results in the same mine may have pointed clearer salt , there are reasons to be optimistic that there is clearer salt in other locations whose radio loss has not yet been precisely measured . for this to be pursued further ,
we believe that developing a portable radar system is a compelling next step so that the radio loss of the salt can be sampled in many locations on a single visit to a mine .
we are grateful to mine manager gordon bull at the north american salt company for supporting our project and his generosity in permitting us to carry out this work in the cote blanche mine on three separate visits .
we would also like to thank scott fountain , robert romero and the other miners at cote blanche that provided us with constant assistance on our visits to make these measurements a success , and for drilling custom boreholes for us which make these measurements unique in the world .
we would also like to thank the high energy physics division of the u.s .
department of energy , the u.k .
particle physics and astronomy research council and the royal society for funding this project .
g. t. zatsepin and v. a. kuzmin , jetp lett . * 4 * , 78 ( 1966 ) .
[ pisma zh .
fiz . * 4 * ,
114 ( 1966 ) ] ; k. greisen , phys .
* 16 * , 748 ( 1966 ) .
r. engel et al . , phys .
d * 64 * , 093010 . for a standard expectation of a neutrino flux from the gzk process , we take the component from electron neutrinos from figure 4 and for the muon neutrino contribution we take the upper curve of figure 9 .
s. barwick , et al .
, j. glaciology * 51 * , 231 ( 2005 ) . g. askaryan , soviet physics jetp - ussr * 14 * ( 2 ) , 441 - 443 ( 1962 ) .
d. saltzberg et al .
* 86 * , 2802 - 2805 ( 2001 ) ; p. gorham et al .
d * 72 * , 023002 ( 2005 ) ; p. gorham et al .
. lett . * 99 * , 171101 ( 2007 ) .
p. gorham , d. saltzberg et al . ,
. meth . * a * 490 , 476 - 491 ( 2002 ) .
r.d.stewart and r.r.unterberger , geophysics * 41 * ( 1 ) , 123 - 132 ( 1976 ) .
m. halbouty , salt domes : gulf region , united states and mexico , gulf publishing co. , houston , tx ( 1967 ) .
b. lock and d.h .
kupfer , salt mines of south louisiana , louisiana geological survey , guidebook series no
. 7 , 2003 . m. cherry , salsa collaboration internal note # 22 ( 2005 ) .
d. wieczorek , salsa collaboration internal note # 21 ( 2005 ) .
a. connolly , a. goodhue , m. cherry , and j. marsh , salsa collaboration internal note # 38 ( 2006 ) .
http://buggs.physics.ucla.edu:8080/salsa/38 a. connolly , a. goodhue , d. saltzberg , salsa collaboration internal note # 39 ( 2006 ) .
r.r.unterberger , geophysical prospecting * 26 * ( 2 ) , 312 - 328 ( 1978 ) . figure [ fig : saltmap ] .
map showing section of 1500 ft .
level of the mine where we made our measurements at boreholes 1 , 2 , and 3 .
the letters and numbers follow the naming system for the corridors used by the mine .
the hatched regions are the salt surrounding the corridors . figure [ fig : timedomain12_13 ] .
pulses that propagated directly between the transmitter and receiver between boreholes 1 and 2 between boreholes 1 and 3 using the mf antennas at a depth of 90 ft .
we plot voltage multiplied by the distance between transmitter and receiver for each pulse , which should be constant in the absence of any attenuation .
figure [ fig : depth ] .
attenuation length at 250 mhz measured with the mf antennas .
the uncertainties are smaller for shorter attenuation lengths since variations in voltage have less of an impact when greater loss is observed .
figure [ fig : avg ] .
average attenuation length , shown as a function of frequency .
the uncertainties shown in this plot are the scatter of the measured attenuation lengths using different antennas and at different depths . figure [ fig : beampattern ] .
measurement of beam pattern while one mf antenna was at 90 ft .
depth in borehole 1 and the other was at various depths in borehole 2 or 3 .
the square markers are measurements taken with the second mf antenna in borehole 2 and the triangular markers with the second mf antenna in borehole 3 .
the star denotes 0 degrees , where the attenuation is by definition 0 db .
lcc + + antenna gain & + frequency & + maximum transmission distance & + + peak power output & 40 dbm & 10 w + loss from transmission ( from friis formula ) & 61 db & @xmath60 + power received ( assuming no attenuation ) & -21 dbm & @xmath61 w + system sensitivity & -100 dbm & @xmath62 w + maximum attenuation allowed & -79 db & @xmath63 + attenuation length & + | we present measurements of the transmission of radio / microwave pulses through salt in the cote blanche salt mine operated by the north american salt company in st .
mary parish , louisiana .
these results are from data taken in the southwestern region of the 1500 ft .
( 457 m ) deep level of the mine on our third and most recent visit to the mine .
we transmitted and received a fast , high - power , broadband pulse from within three vertical boreholes that were drilled to depths of 100 ft .
( 30 m ) and 200 ft .
below the 1500 ft .
level using three different pairs of dipole antennas whose bandwidths span 125 to 900 mhz . by measuring the relative strength of the received pulses between boreholes with separations of 50 m and 169 m , we deduce the attenuation of the signal attributed to the salt medium .
we fit the frequency dependence of the attenuation to a power law and find the best fit field attenuation lengths to be @xmath0 m at 150 mhz , @xmath1 m at 300 mhz , and @xmath2 m at 800 mhz .
this is the most precise measurement of radio attenuation in a natural salt formation to date .
we assess the implications of this measurement for a future neutrino detector in salt . , , , , salt , radio , microwave , transmission , attenuation , neutrino 42.25.bs , 91.55.de , 93.85.fg , 95.85.ry |
aspects of the @xmath0 factorization theorem @xcite in perturbative qcd have been investigated intensively .
one of the important issues is about the derivation of a @xmath0-dependent hard kernel at subleading level , which is defined as the difference between qcd diagrams and effective diagrams for transverse - momentum - dependent ( tmd ) hadron wave functions . we have explained that partons in both sets of diagrams should remain off mass shell by @xmath1 in the @xmath0 factorization theorem @xcite .
the same statement has been made in the application of the @xmath0 factorization theorem to inclusive processes such as prompt photon production @xcite .
the off - shellness of partons may cause concern of the gauge invariance @xcite .
however , we have shown that the gauge dependence cancels between the above two sets of diagrams , and a @xmath0-dependent hard kernel is gauge invariant @xcite . following this prescription , the next - to - leading - order ( nlo ) correction to the pion transition ( electromagnetic ) form factor associated with the process @xmath2 has been calculated at leading twist , i.e. , twist 2 @xcite . here
we shall study the power - suppressed three - parton contribution to the pion electromagnetic form factor in the @xmath0 factorization theorem .
the three - parton contribution in the collinear factorization theorem @xcite to a simpler process , the @xmath3 meson transition form factor , has been evaluated recently @xcite .
we shall first demonstrate the gauge invariance of the three - parton contribution to the pion electromagnetic form factor in the @xmath0 factorization theorem .
there are two sources of gauge dependence for this power correction @xcite : the first source is proportional to parton transverse momenta from the two - parton fock state . the corresponding hadronic matrix element is written as @xmath4 where @xmath5 is the coordinate of the anti - quark field @xmath6 , and @xmath7 represents a combination of gamma matrices .
the second source is associated with the three - parton fock state with an additional valence gluon .
the corresponding matrix element is given by @xmath8 with a color matrix @xmath9 , and the gluon field @xmath10 at the coordinate @xmath11 .
the gauge dependences from the above two sources cancel each other , when eqs .
( [ d3 ] ) and ( [ a3 ] ) are combined to form the gauge - invariant matrix element @xmath12 .
the cancellation of the gauge dependences is similar to that occurring in the collinear factorization theorem @xcite . in the @xmath0 factorization theorem
we just keep the transverse momentum dependence in denominators of particle propagators @xcite .
hence , it is natural that the gauge dependence disappears at higher twists in the same way as in the collinear factorization theorem .
our formalism implies that contributions proportional to transverse momenta in numerators of hard kernels must be combined with contributions from three - parton fock states in order to guarantee gauge invariance .
therefore , there is concern on the study of the pion transition form factor in @xcite , where only the former was included .
it has been pointed out explicitly that the leading - order ( lo ) hard kernel for the pion electromagnetic form factor becomes gauge - dependent , if one simply considers parton transverse momenta in numerators @xcite . however , the contribution from the three - parton fock state was still missing in @xcite , such that the false postulation on the gauge dependence of the @xmath0 factorization theorem was made . since both the
initial- and final - state pions involve higher - twist matrix elements like that in eq .
( [ t3 ] ) , the three - parton contribution to the pion electromagnetic form factor is suppressed at least by @xmath13 , @xmath14 being the momentum transfer squared .
after examining the gauge invariance , we calculate the @xmath0-dependent hard kernel for the three - parton - to - three - parton scattering in the feynman gauge , and convolute it with the three - parton pion wave functions .
we observe that the diagrams with a four - gluon vertex dominate this power correction .
it will be shown that the three - parton contribution is only about 5% of the sum of those which have been investigated before , including the lo twist-2 , nlo twist-2 , and lo two - parton twist-3 ones @xcite .
that is , the three - parton contribution is not crucial for accommodating experimental data of the pion electromagnetic form factor . at the same power of @xmath13
, one should also take into account the scattering of two ( four ) partons into four ( two ) partons in principle .
this piece has been analyzed in light - cone sum rules @xcite , and found to be less important than other contributions . with this work ,
we conclude that the chirally enhanced two - parton twist-3 correction is the most important @xmath13 correction to the pion electromagnetic form factor . in sec .
ii we verify the gauge invariance of the lo three - parton contribution to the pion electromagnetic form factor by combining the gauge - dependent hard kernels corresponding to eqs .
( [ d3 ] ) and ( [ a3 ] ) .
the @xmath0-dependent hard kernel is then derived from the three - parton - to - three - parton scattering diagrams , and convoluted with the three - parton pion wave functions numerically in sec .
section iv is the conclusion .
detailed calculations of the gauge - dependent hard kernels corresponding to eq .
( [ a3 ] ) are presented in appendix a , and the expressions of the three - parton - to - three - parton hard kernels are collected in appendix b.
consider the pion electromagnetic form factor involved in the process @xmath15 , whose lo diagrams are displayed in fig .
[ fig1 ] . the momentum @xmath16 of the initial - state pion and @xmath17 of the final - state pion are parameterized as @xmath18 with @xmath19 , @xmath20 being the virtual photon momentum .
the gluon propagator of momentum @xmath21 is written as @xmath22 in the covariant gauge , where the parameter @xmath23 will be used to identify sources of gauge dependence .
we assume that the anti - quarks in the initial- and final - state pions , represented by lower fermion line , carry the parton momenta @xmath24 respectively , @xmath25 and @xmath26 being the momentum fractions .
it is understood that the components @xmath27 and @xmath28 have been dropped in hard kernels , and integrated out of the tmd pion wave functions
. represents the virtual photon vertex.,height=124 ] we employ the fierz identity @xmath29 to factorize the fermion flow , where @xmath30 denotes the @xmath31 identity matrix .
the structure @xmath32 in the above identity contributes at twist 2 and higher twists , and @xmath33 and @xmath34 contribute at twist 3 and higher twists at two - parton level .
the matrix @xmath7 in eqs .
( [ d3 ] ) and ( [ a3 ] ) can pick up one of the above structures , among which we focus on the first one @xmath35 as an example below .
we also insert the identity @xmath36 to factorize the color flow , where @xmath37 is the number of colors , @xmath30 denotes the @xmath38 identity matrix , and @xmath39 is a color matrix . the first ( second ) term in eq .
( [ col ] ) will be associated with a color - singlet ( color - octet ) state of the valence quark and anti - quark .
the first source of gauge dependence is extracted from the diagrams in fig .
[ fig1 ] , where the quark and anti - quark pair forms a color - singlet state . combining the decompositions in eqs .
( [ fie ] ) and ( [ col ] ) , we sandwich fig .
[ fig1 ] with the structures @xmath40 from the initial and final states , respectively , where the subscripts @xmath41 and @xmath42 can take arbitrary components .
the lo hard kernel from fig .
[ fig1](a ) contains the gauge - dependent piece @xmath43 } { ( p_1-k_2)^2(k_1-k_2)^2}\frac{(k_1-k_2)_\sigma ( k_1-k_2)_\nu}{(k_1-k_2)^2},\label{hlam}\end{aligned}\ ] ] with @xmath44 being a color factor .
the above expression diminishes with the substitution @xmath45 , @xmath46 , @xmath47 ( proportional to @xmath48 ) , and @xmath49 ( proportional to @xmath50 ) into the numerator , implying that eq .
( [ hlam ] ) does not contribute at leading power in the @xmath0 factorization theorem . to obtain the gauge - dependent hard kernel from eq .
( [ hlam ] ) at @xmath13 , we insert the identity @xmath51 . it can be shown that the contribution from the @xmath52 term is cancelled by the corresponding one in fig .
[ fig1](b ) .
the second term , with @xmath53 being the momentum of the incoming valence quark , corresponds to the matrix element with the derivative of the quark field in the initial - state pion .
this term can be picked up by differentiating eq .
( [ hlam ] ) with respect to @xmath54 .
once eq .
( [ d3 ] ) for the initial - state pion is identified , we pick up the @xmath55 term in @xmath56 via differentiation , which corresponds to the derivative of the valence anti - quark field in the final - state pion . note that denominators of particle propagators , depending on @xmath54 and @xmath57 , will be differentiated too . however , their differentiation gives rise to even higher - twist matrix elements , and can be neglected .
we then extract @xmath58 } { ( p_1-k_2)^2(k_1-k_2)^4},\label{tt}\end{aligned}\ ] ] associated with @xmath59 and the similar matrix element for the final - state pion .
the lo hard kernel from fig .
[ fig1](b ) contains @xmath60}{(p_2-k_1)^2(k_1-k_2)^2}\frac{(k_1-k_2)_\sigma ( k_1-k_2)_\nu}{(k_1-k_2)^2},\label{hlbm}\end{aligned}\ ] ] whose differentiation with respect to @xmath61 and @xmath62 leads to @xmath63 } { ( p_2-k_1)^2(k_1-k_2)^4}.\label{ttb}\end{aligned}\ ] ] this hard kernel corresponds to the matrix element of the initial - state pion with the derivative of the anti - quark field @xmath64 .
equations ( [ tt ] ) and ( [ ttb ] ) represent the gauge dependence in lo two - parton - to - two - parton scattering at power of @xmath13 , which was also observed in @xcite .
, @xmath65 , @xmath66 denote the attachments of the additional valence gluon.,title="fig:",height=105 ] , @xmath65 , @xmath66 denote the attachments of the additional valence gluon.,title="fig:",height=105 ] the second source of gauge dependence arises from fig . [ fig2 ] , where three partons appear in the initial state as indicated by eq .
( [ a3 ] ) .
all possible attachments of the additional valence gluon to the lines other than the valence quark and anti - quark in the initial state are labeled by letters @xmath67 , @xmath65 , @xmath66 . in principle , the diagrams with the attachments to the valence quark and anti - quark in the initial state should be included in order to respect @xmath68 gauge symmetry for the electromagnetic interaction gauge symmetry . ] .
these diagrams contribute to higher gegenbauer terms in the two - parton twist-3 pion distribution amplitudes .
equations of motion can then be constructed to relate the coefficients of the higher gegenbauer terms in the two - parton twist-3 and three - parton twist-3 pion distribution amplitudes @xcite .
hence , one should pay attention to the consistency between the models for these two sets of distribution amplitudes in a numerical analysis .
we shall adopt the non - asymptotic models for both sets of distribution amplitudes in sec .
iii , when estimating the importance of the three - parton contribution relative to other two - parton contributions in the pion electromagnetic form factor .
according to eq .
( [ col ] ) , we absorb a color matrix @xmath39 and the coupling constant @xmath69 associated with an attachment into the matrix element for the initial - state pion , and another @xmath39 goes into the evaluation of gauge - dependent hard kernels . for example , the color factor corresponding to the attachment @xmath67 of the valence gluon to the virtual quark line is given by @xmath70 = -\frac{1}{4n_c } \delta^{ac},\label{ca}\end{aligned}\ ] ] where the color matrix @xmath9 ( @xmath71 ) comes from the valence ( hard ) gluon vertex .
after summing over @xmath72 , the tensor @xmath73 sets @xmath74 in the matrix element for the initial - state pion , leading to eq .
( [ a3 ] ) . including the coefficient 2 in eq .
( [ col ] ) , we adopt the structure @xmath75 for the initial state in the calculation of fig .
[ fig2 ] , whose details can be found in appendix a. the results are collected as follows : @xmath76 } { ( p_1-k_2)^2(k_1-k_2)^4},\label{aa}\\ h_{bt}^{\lambda } & = & ieg^2\lambda \frac{1}{32}\frac{tr[\gamma^\beta\gamma_5 \gamma_{\beta}\gamma_\mu ( \not p_1-x_2\not p_2)\gamma^\alpha \gamma_{\alpha}\gamma_5]}{(p_1-k_2)^2(k_1-k_2)^4},\label{b}\\ h^{\lambda}_{ct}&=&0,\label{c}\\ h^{\lambda}_{dt}&=&-ieg^2\lambda\frac{1}{32n_c^2 } \frac{tr[\gamma^\beta \gamma_5 \gamma_{\beta}\gamma^\alpha(\not p_2-x_2\not p_2-y_1\not p_1)\gamma_\mu \gamma_{\alpha}\gamma_5]}{(p_2-k_2-l_1)^2(k_1-k_2)^4},\label{d}\\ h^{\lambda}_{et}&=&-ieg^2\lambda\frac{c_f}{16n_c } \frac{tr[\gamma^\beta \gamma_5\gamma_{\beta } \gamma^\alpha ( \not p_2- x_1\not p_1-y_1\not p_1 ) \gamma_\mu \gamma_{\alpha}\gamma_5]}{(p_2-k_1-l_1)^2(k_1-k_2)^4},\label{e}\\ h^{\lambda}_{ft}&=&ieg^2\lambda\frac{1}{32n_c^2 } \frac{tr[\gamma^\beta \gamma_5\gamma_{\beta}\gamma^\alpha ( \not p_2- x_2\not p_2-y_1\not p_1)\gamma_\mu \gamma_{\alpha}\gamma_5 ] } { ( p_2-k_2-l_1)^2(k_1-k_2)^4}\nonumber\\ & & -ieg^2\lambda\frac{1}{32n_c^2 } \frac{tr[\gamma^\beta \gamma_5\gamma_{\beta } \gamma^\alpha(\not p_2-x_1\not p_1-y_1\not p_1)\gamma_\mu \gamma_{\alpha}\gamma_5]}{(p_2-k_1-l_1)^2(k_1-k_2)^4},\label{f}\\ h^{\lambda}_{gt } & = & ieg^2\lambda \frac{1}{32}\frac{tr[\gamma^\beta \gamma_5\gamma_{\beta}\gamma^\alpha ( \not p_2-x_1\not p_1-y_1\not p_1)\gamma_\mu \gamma_{\alpha}\gamma_5]}{(p_2-k_1-l_1)^2(k_1-k_2)^4}\nonumber\\ & & + ieg^2 \lambda\frac{1}{32}\frac{tr[\gamma^\alpha\gamma_5 \gamma_{\beta } \gamma^\beta(\not p_2-x_1\not p_1-y_1\not p_1)\gamma_\mu \gamma_{\alpha}\gamma_5]}{(p_2-k_1-l_1)^2(k_1+l_1-k_2)^4},\label{g}\\ h^{\lambda}_{ht } & = & -ieg^2\lambda\frac{1}{32n_c^2 } \frac{tr [ \gamma^\alpha\gamma_5\gamma_{\beta}\gamma^\beta(\not p_2-x_1\not p_1-y_1\not p_1)\gamma_\mu \gamma_{\alpha}\gamma_5]}{(p_2-k_1-l_1)^2 ( k_1+l_1-k_2)^4},\label{h}\end{aligned}\ ] ] with the gluon momentum fraction @xmath77 . summing the above expressions , we arrive at @xmath78 with @xmath79 the contributions from the attachments @xmath67 and @xmath65 are added into @xmath80 with the desired color factor @xmath44 .
the contribution from the attachment @xmath81 and the first term in the attachment @xmath82 cancels each other . the second term from the attachment @xmath82 and the first term from the attachment @xmath83
are combined into the expression with the color factor @xmath44 , which then cancels the contribution from the attachment @xmath84 .
the second term of @xmath83 and the contribution from the attachment @xmath85 are added into @xmath86 .
note that @xmath80 does not depend on the valence gluon momentum @xmath87 , which can then be integrated out of the matrix element , giving @xmath88 , corresponds to the matrix element @xmath89 .
because of the symmetry under the exchange of the initial- and final - state kinematic variables , the gauge - dependent hard kernels with three partons from the final state are written as @xmath90 where @xmath91 is the momentum carried by the outgoing valence gluon . at last ,
we extract the gauge dependence from the three - parton - to - three - parton diagrams . in this case the indices @xmath41 and @xmath42 , associated with the initial and final valence gluons , respectively , must be carried by gluon vertices , instead of by parton momenta . focusing on the gauge - dependent piece
, we can apply the ward identity to hard gluons .
it is easy to find that fig .
[ figc ] , where the two valence gluons scatter via a three - gluon vertex , does not contribute : if the gauge dependence arises from the lower hard gluon , the results , being proportional to the parton momenta after applying the ward identity , should be dropped .
if the gauge dependence arises from the upper hard gluon , the ward identity diminishes the amplitude for a similar reason .
it is also easy to see that fig .
[ fig9](a ) with a four - gluon vertex does not contribute to a gauge - dependent hard kernel .
if both the valence gluons attach to the quark line , the gauge - dependent contribution vanishes because of the ward identity applied to the anti - quark line .
if both the valence gluons attach to the anti - quark line , the gauge - dependent contribution vanishes too . 0.5 cm 0.5 cm the other diagrams are classified into several sets , which are formed by all possible attachments of a hard gluon as displayed in figs .
[ fig3](a)-[fig3](l ) .
cancellation occurs in each set of diagrams after applying the ward identity . neglecting those pieces proportional to the parton momenta
, we have @xmath92 a finite gauge - dependent contribution comes only from figs .
[ fig3](i)-[fig3](l ) , given by @xmath93 due to the same expressions of @xmath94 and @xmath80 , their corresponding matrix elements are combined into @xmath95 .
that is , the gauge invariance holds , when the contributions from figs . [ fig1 ] and [ fig2 ] are combined .
the combination of the matrix elements for @xmath96 and @xmath97 also vanishes according to eq .
( [ eq ] ) .
a similar reasoning applies to the gauge - dependent hard kernels @xmath98 and @xmath99 and to @xmath86 and @xmath100 : the combination of their matrix elements vanishes due to the equation of the motion for the quark field in the final - state pion .
this observation completes the proof of the gauge invariance of the lo three - parton contribution to the pion electromagnetic form factor at power of @xmath13 in the @xmath0 factorization theorem .
the extension of the proof to all orders can follow the steps outlined in @xcite . note that our proof applies to the collinear factorization theorem too : simply neglecting transverse momenta in denominators , one can show the gauge invariance of the three - parton contribution to the pion electromagnetic form factor in the collinear factorization theorem .
in this section we calculate the three - parton contribution to the pion electromagnetic form factor .
start with the gauge - invariant twist-3 matrix element @xmath101 where the subscript @xmath41 is associated with the vertex the valence gluon attaches to .
the power behavior will not be changed , and the gauge invariance will not be broken by inserting another covariant derivative @xmath102 .
we then exchange @xmath102 and @xmath103 , take the difference of @xmath104 and @xmath105 , and apply the identity @xmath106=-igg^+_{\hspace{0.2cm}\alpha}$ ] .
it is then equivalent to employ the following alternative matrix element @xcite , which defines the three - parton twist-3 pion wave function @xmath107 , @xmath108 with the chiral scale @xmath109 , @xmath110 , @xmath111 , and @xmath112 being the pion , @xmath113 quark and @xmath114 quark masses , respectively .
the operators with other spin structures contribute at higher twists : for example , the operator @xmath115 gives a three - parton twist-4 contribution , and @xmath116 does not contribute @xcite . with eq .
( [ gi3 ] ) , one may verify the gauge invariance of a hard kernel in the @xmath0 factorization theorem by demonstrating the cancellation between the gauge - dependent contributions from the operators @xmath117 and @xmath118 @xcite .
below we derive the hard kernels from the three - parton - to - three - parton diagrams corresponding to eq .
( [ gi3 ] ) in the feynman gauge ( @xmath119 ) . choosing this gauge ,
the operator @xmath118 does not contribute , so only @xmath117 is relevant .
the three momenta @xmath120 , and @xmath87 are assigned to the initial - state quark , antiquark , and gluon , respectively , and @xmath121 , and @xmath91 to the final - state quark , anti - quark , and gluon , respectively .
we have the structures for the initial- and final - state pions @xmath122 where the gluon momentum fraction @xmath123 is defined by @xmath124 , and the gamma matrix @xmath125 involves only transverse components .
there are totally 196 diagrams for the three - parton - to - three - parton scattering , which can be divided into four categories .
category a contains 20 quark - gluon configurations , in which neither of the valence gluons attaches to the hard gluon line .
each configuration allows 6 different attachments for the photon line .
we further divide this category into two groups as shown in fig .
[ fig6 ] , where both valence gluons attach to the same quark or anti - quark line , and in fig .
[ fig7 ] , where one valence gluon attaches to the quark line and another to the anti - quark line .
only the diagrams giving nonvanishing contributions are displayed . as observed in appendix b , the amplitudes with both
the valence gluons attaching to the quark line are power - suppressed .
category b contains 8 quark - gluon configurations , in which one of the valence gluons attaches to the hard gluon line .
each configuration allows 5 different attachments for the photon line , among which those with nonvanishing contributions are displayed in fig .
category c contains 4 quark - gluon configurations , where both the valence gluons are connected to the hard gluon .
each configuration allows 4 different attachments for the photon line , among which those with nonvanishing contributions are displayed in fig .
category d contains 4 quark - gluon configurations , where the two valence gluons scatter via a three - gluon vertex as shown in fig .
each configuration allows 5 different attachments for the photon line .
since this category of diagrams does not contribute , we shall not discuss them further .
besides , when a valence gluon attaches to a valence quark , the diagram should be regarded as being from an effective two - parton fock state , and will not be calculated .
we extract the hard kernels proportional to the final - state momentum @xmath126 .
the hard kernels proportional to @xmath127 can be obtained by exchanging the kinetic variables of the initial- and final - state pions . adopting the electric charge @xmath128 , instead of the quark charge @xmath129 or @xmath130 , we have taken into account the diagrams with the virtual photon attaching to the anti - quark line .
figure [ fig9](a ) with a four - gluon vertex gives the dominant three - parton contribution @xmath131}{(p_1-k_2-l_2)^2(k_1-k_2)^2(k_1-k_2+l_1-l_2)^2 } ( g_{\alpha \nu } g_{\beta \sigma } + g_{\alpha \sigma } g_{\beta \nu}-2 g_{\alpha\beta } g_{\sigma\nu})\nonumber\\ & = & -i eg^2\frac{f_\pi^2 m_0 ^ 2}{16y_1y_2 } \frac { n_c^2 } { n_c^2 - 1 } \frac{tr [ \not p_2\gamma_\mu \not p_2\not p_1 ] } { ( p_1-p_2)^2(k_1-k_2+l_1-l_2)^2(k_1-k_2)^2}.\label{aa1}\end{aligned}\ ] ] to arrive at the second line , we have made an approximation according to the power counting @xmath132 @xcite , under which the tmd term in the following denominator is neglected , @xmath133 the expressions for other three - parton - to - three - parton diagrams are collected in appendix b. since the sudakov factor for exclusive qcd processes was derived in the space of impact parameters @xcite , we fourier transform eq .
( [ aa1 ] ) . the @xmath0 factorization formula for the pion electromagnetic pion form factor from fig .
[ fig9](a ) is then written as @xmath134 where the three - parton pion distribution amplitude @xmath135 corresponds to @xmath107 in eq .
( [ gi3 ] ) in the space of momentum fractions .
the functions @xmath136 , arising from the fourier transformation of the tmd denominators @xmath137 and @xmath138 in eq .
( [ aa1 ] ) , are defined by @xmath139,\label{kt}\end{aligned}\ ] ] in which @xmath140 is the modified bessel function , and the explicit expression of the sudakov exponent @xmath141 is referred to @xcite .
we have kept only the most effective piece of the sudakov evolution in the small @xmath142 region , that results from the gluon exchanges between the energetic valence quark and the wilson line associated with it . because the sudakov factor @xmath143 $ ] diminishes at @xmath144 , with the qcd scale @xmath145 gev ,
the upper bound of the integration variable has been set to @xmath146 in eq .
( [ kt ] ) . for order - of - magnitude estimate and for demonstrating the smallness of the three - parton contribution
, we do not consider the renormalization - group evolution from the low scale , at which @xmath147 is defined , to the scale of the hard kernel .
the coupling constant is also assumed to be a constant @xmath148 .
dependence of the three - parton contribution to the pion electromagnetic form factor .
the lo twist-2 , nlo twist-2 , and lo two - parton twist-3 contributions are quoted from fig . 8 in @xcite , which were derived using the non - asymptotic models for the pion distribution amplitudes.,height=566 ] fourier transforming the hard kernels in appendix b into the impact - parameter space , we construct the corresponding @xmath0 factorization formulas similar to eq .
( [ f3 ] ) .
we then add all the contributions to the pion electromagnetic form factor , and employ the model of the three - parton twist-3 pion distribution amplitude @xcite @xmath149,\end{aligned}\ ] ] for a numerical analysis , with the parameters @xmath150 and @xmath151 , and @xmath152 being the momentum fraction of the valence quark .
the total three - parton contribution to @xmath153 , @xmath154 being the pion electromagnetic form fact , is displayed in fig .
the curve exhibits a decrease in @xmath14 ( though not obvious in the figure ) compared to the lo and nlo twist-2 contributions , indicating that this contribution is power - suppressed .
it is only about 5% of the sum of those evaluated in @xcite , including the lo twist-2 , nlo twist-2 , and lo two - parton twist-3 pieces .
that is , the three - parton contribution is not crucial for accommodating the experimental data of @xmath153 @xcite in the whole accessible range of @xmath14 up to 10 gev@xmath155 .
the only important subleading contribution to the pion electromagnetic form factor that have been investigated so far comes from the chirally enhanced two - parton twist-3 one .
in this paper we have applied the @xmath0 factorization theorem to the study of the power - suppressed three - parton contribution to the pion electromagnetic form factor .
it was demonstrated that the gauge invariance of the @xmath0-dependent hard kernel holds for this power correction : the gauge dependence proportional to parton transverse momenta from the two - parton fock state and the gauge dependence associated with the three - parton fock state cancel each other .
we have calculated the three - parton - to - three - parton hard kernel at lo , and found that the three - parton contribution is about 5% of the sum of the lo twist-2 , nlo twist-2 , and lo two - parton twist-3 ones in the whole range of experimentally accessible @xmath14 .
our analysis shows that the power expansion for this exclusive process might be reliable in the @xmath0 factorization theorem . at the same power of @xmath13
, the two - parton twist-4 contribution should be taken into account , which has been studied in the framework of light - cone sum rules @xcite .
we shall calculate this correction in the @xmath0 factorization theorem in the future , which involves a twist-2 distribution amplitude from one side and a two - parton twist-4 distribution amplitude from the other side .
we shall also extend our framework to exclusive @xmath65 meson decays , for which three - parton contributions have been analyzed in light - cone sum rules , @xcite , in the qcd ( collinear ) factorization @xcite and in the soft - collinear effective theory @xcite .
we thank z.t .
wei for his suggestion , which initiates this project .
we also thank v. braun , s. brodsky , j.p . ma and y.m .
wang for useful discussions .
this work was supported in part by the national science council of r.o.c . under grant
nsc-98 - 2112-m-001 - 015-my3 , and by the national center for theoretical sciences .
in this appendix we present the detailed derivation of the gauge - dependent hard kernels from fig .
[ fig2 ] . the attachment @xmath67 contains , with the color factor in eq .
( [ ca ] ) , the gauge - dependent piece @xmath156}{(p_1-k_2)^2(p_1-l_1-k_2)^2(k_1-k_2)^4}.\label{at1}\end{aligned}\ ] ] inserting the identity @xmath157 into eq .
( [ at1 ] ) , we obtain @xmath158}{(p_1-k_2)^2(k_1-k_2)^4}\nonumber\\ & & -\frac{tr[(\not k_1-\not k_2 ) \gamma_5 \gamma_{\beta}\gamma_\mu ( \not p_1-\not k_2)\gamma^\alpha(\not p_1-\not l_1-\not k_2)(\not p_1-\not k_1-\not l_1 ) \gamma_{\alpha}\gamma_5 ] } { ( p_1-k_2)^2(p_1-l_1-k_2)^2(k_1-k_2)^4}\bigg\}.\label{13}\end{aligned}\ ] ] the second term , proportional to the momentum @xmath159 of the incoming valence quark , should be dropped , since there is the valence gluon @xmath160 from the initial state already .
taking the derivative of the first term with respect to @xmath62 , and then substituting @xmath45 and @xmath46 into the numerator , we have eq . ( [ aa ] ) .
the diagram with the attachment @xmath65 of the valence gluon to the hard gluon line produces the gauge - dependent hard kernel @xmath161}{(p_1-k_2)^2 } tr(t^dt^bt^c)\gamma^{dba}_{\delta\nu\alpha}\nonumber\\ & & \times\bigg[\lambda\frac{g^{\delta\delta'}}{(k_1-k_2)^2 } \frac{(k_1-k_2+l_1)^\nu(k_1-k_2+l_1)^{\nu'}}{(k_1-k_2+l_1)^4}+\lambda \frac{(k_1-k_2)^\delta(k_1-k_2)^{\delta'}}{(k_1-k_2)^4 } \frac{g^{\nu\nu'}}{(k_1-k_2+l_1)^2}\nonumber\\ & & -\lambda^2\frac{(k_1-k_2)^\delta(k_1-k_2)^{\delta'}}{(k_1-k_2)^4 } \frac{(k_1-k_2+l_1)^\nu(k_1-k_2+l_1)^{\nu'}}{(k_1-k_2+l_1)^4}\bigg],\end{aligned}\ ] ] with the triple - gluon vertex , @xmath162 , \label{tg3}\end{aligned}\ ] ] @xmath163 being a antisymmetric tensor .
using the identity @xmath164 @xmath165 being a symmetric tensor , the above amplitude becomes @xmath166}{(p_1-k_2)^2 } \nonumber\\ & & \times [ g_{\alpha\nu}(2l_1+k_1-k_2)_\delta + g_{\nu\delta}(2k_2 - 2k_1-l_1)_\alpha + g_{\delta\alpha}(k_1-k_2-l_1)_\nu]\nonumber\\ & & \times\bigg[\lambda\frac{g^{\delta\delta'}}{(k_1-k_2)^2 } \frac{(k_1-k_2+l_1)^\nu(k_1-k_2+l_1)^{\nu'}}{(k_1-k_2+l_1)^4}+\lambda \frac{(k_1-k_2)^\delta(k_1-k_2)^{\delta'}}{(k_1-k_2)^4 } \frac{g^{\nu\nu'}}{(k_1-k_2+l_1)^2}\nonumber\\ & & -\lambda^2\frac{(k_1-k_2)^\delta(k_1-k_2)^{\delta'}}{(k_1-k_2)^4 } \frac{(k_1-k_2+l_1)^\nu(k_1-k_2+l_1)^{\nu'}}{(k_1-k_2+l_1)^4}\bigg ] .
\label{hb}\end{aligned}\ ] ] the @xmath167 term leads to @xmath168}{(p_1-k_2)^2(k_1-k_2)^4(k_1-k_2+l_1)^4}\nonumber\\ & & \times\big[(k_1-k_2)\cdot(2l_1+k_1-k_2)(k_1-k_2+l_1)^\alpha+ ( k_1-k_2)\cdot(k_1-k_2+l_1)(2k_2 - 2k_1-l_1)^\alpha\nonumber\\ & & + ( k_1-k_2-l_1)\cdot(k_1-k_2+l_1)(k_1-k_2)^\alpha\big].\end{aligned}\ ] ] inserting the identity @xmath169 , it is easy to see that the first term of the identity gives an expression which is cancelled by the corresponding one from the attachment @xmath83 .
the second term , being proportional to the momentum @xmath159 of the incoming valence quark , should be neglected .
we then consider the @xmath23 terms . the first @xmath23 term in eq .
( [ hb ] ) gives @xmath170 } { ( p_1-k_2)^2(k_1-k_2)^2(k_1-k_2+l_1)^4}(k_1-k_2+l_1)^\alpha \nonumber\\ -ieg^2 \frac{1}{32}\lambda\frac{tr[(\not k_1-\not k_2+\not l_1)\gamma_5\gamma_{\beta}\gamma_\mu ( \not p_1-\not k_2)(\not k_1-\not k_2+\not l_1)\gamma_{\alpha}\gamma_5 ] } { ( p_1-k_2)^2(k_1-k_2)^2(k_1-k_2+l_1)^4}(2k_2 - 2k_1-l_1)^\alpha \nonumber\\ -ieg^2 \frac{1}{32}\lambda\frac{tr[\gamma^\alpha\gamma_5 \gamma_{\beta}\gamma_\mu ( \not p_1-\not k_2)(\not k_1-\not k_2+\not l_1)\gamma_{\alpha}\gamma_5 ] } { ( p_1-k_2)^2(k_1-k_2)^2(k_1-k_2+l_1)^4}(k_1-k_2+l_1)\cdot(k_1-k_2-l_1),\end{aligned}\ ] ] which are all negligible for the same reason as for the @xmath167 term . hence , @xmath171 receives a contribution only from the second @xmath23 term , @xmath172 } { ( p_1-k_2)^2(k_1-k_2)^4(k_1-k_2+l_1)^2 } ( k_1-k_2)\cdot(2l_1+k_1-k_2)\nonumber\\ & = & -ieg^2 \frac{1}{32}\lambda\frac{tr[(\not k_1-\not k_2)\gamma_5 \gamma_{\beta}\gamma_\mu ( \not p_1-\not k_2)\gamma^\alpha \gamma_{\alpha}\gamma_5]}{(p_1-k_2)^2(k_1-k_2)^4}.\end{aligned}\ ] ] to arrive at the second line , the higher - power term @xmath173 has been added , so we have @xmath174 .
the differentiation of the above expression with respect to @xmath62 leads to eq .
( [ b ] ) . the diagrams with other attachments in fig .
[ fig2 ] can be calculated in a similar way , so the detail will not be presented here .
in this appendix we collect the expressions of the three - parton - to - three - parton hard kernels for the pion electromagnetic form factor in the feynman gauge .
start with category a defined in sec .
iii . when the attachments of the two valence gluons are arranged in the way that the hard gluon vertices sandwich the spin structures associated with the pions , the contribution diminishes because of @xmath175 or @xmath176 .
the nonvanishing amplitudes come from figs .
[ fig6 ] and [ fig7 ] , which are written as @xmath177 it is observed that the results from figs .
[ fig6](a ) and [ fig6](c ) are suppressed by a power of @xmath178 compared to fig .
[ fig6](b ) .
that is , when both the valence gluons attach to the quark line , the contribution is power - suppressed .
the hard kernels from fig .
[ fig7 ] are of the same power as eq .
( [ 7b ] ) .
figure [ fig8 ] from category b gives the hard kernels @xmath179}{(p_2-k_1-l_1)^2(k_1-k_2+l_1)^2(k_1-k_2)^2(p_1-k_1-l_1-p_2)^2 } , \\
h_{8b}&= & \frac{i}{2}eg^2\frac{1}{n_c^2 - 1}\frac{tr[\not p_2\gamma_\mu \not p_2 ( \not k_1 - \not l_1)\not p_2\not p_1]}{(p_1-p_2)^2(k_1-k_2+l_1)^2(k_1-k_2)^2 ( p_1-k_1-l_1-p_2)^2 } , \\
h_{8c}&= & \frac{i}{2}eg^2\frac{tr ( \not p_2\gamma_\mu\not p_2\not p_1 ) } { ( p_1-p_2)^2(k_1-k_2+l_1)^2(k_1-k_2)^2 } , \\ h_{8d}&= & -\frac{i}{2}eg^2 \frac{1}{n_c^2 - 1}\frac{tr[\not p_2\gamma_\mu\not p_2\not p_1(\not k_2 + 2\not l_2)\not l_1 ] } { ( p_1-p_2)^2(k_1-k_2+l_1)^2(l_1-k_2)^2(k_1-k_2+l_1-l_2)^2 } , \\
h_{8e}&= & -\frac{i}{2}eg^2\frac{1}{n_c^2 - 1}\frac{tr[\not p_2\gamma_\mu\not p_2 \not p_1\not l_2(\not k_1 + 2\not l_1 ) ] } { ( p_1-p_2)^2(k_1-k_2+l_1-l_2)^2(k_1-l_2)^2 ( k_1-k_2-l_2)^2 } , \\ h_{8f}&= & -\frac{i}{2}eg^2\frac{1}{n_c^2 - 1}\frac{tr [ \not
p_2 \gamma_\mu\not p_2\not p_1(\not k_2 - \not l_2)\not p_1 ] } { ( p_1-p_2)^2(k_1-k_2)^2 ( k_1-k_2-l_2)^2(p_1-k_2-l_2)^2}.\end{aligned}\ ] ] 99 s. catani , m. ciafaloni and f. hautmann , phys .
b * 242 * , 97 ( 1990 ) ; nucl . phys . *
b366 * , 135 ( 1991 ) .
collins and r.k .
ellis , nucl .
* b360 * , 3 ( 1991 ) .
levin , m.g .
ryskin , yu.m .
shabelskii , and a.g .
shuvaev , sov .
. phys . * 53 * , 657 ( 1991 ) .
j. botts and g. sterman , nucl .
b225 * , 62 ( 1989 ) .
li and g. sterman , nucl .
b381 * , 129 ( 1992 ) . t. huang and q.x .
shen , z. phys .
c * 50 * , 139 ( 1991 ) ; j.p .
ralston and b. pire , phys .
* 65 * , 2343 ( 1990 ) . m. nagashima and h - n .
li , phys .
d * 67 * , 034001 ( 2003 ) .
lipatov , m.a .
malyshev , and n.p .
zotov , arxiv:1102.1134 [ hep - ph ] .
f. feng , j.p .
ma and q. wang , phys . lett .
b * 674 * , 176 ( 2009 ) .
f. feng , j.p .
ma and q. wang , arxiv:1103.1706 [ hep - ph ] .
li and s. mishima , phys .
b * 674 * , 182 ( 2009 ) .
li and s. mishima , phys .
d * 80 * , 074024 ( 2009 ) .
s. nandi and h - n .
li , phys .
d * 76 * , 034008 ( 2007 ) .
wang , y.l . shen , and h. zou , phys .
d * 83 * , 054029 ( 2011 ) .
lepage and s.j .
brodsky , phys .
d * 22 * , 2157 ( 1980 ) .
anikin , d.yu .
ivanov , b. pire , l. szymanowski , and s. wallon , nucl .
* b828 * , 1 ( 2010 ) .
qiu , phys .
d * 42 * , 30 ( 1990 ) ; j. qiu and g. sterman , nucl .
b378 * , 52 ( 1992 ) .
f.g . cao , t. huang , and b.q .
ma , phys .
d * 53 * , 6582 ( 1996 ) .
j. bijnens and a. khodjamirian , eur .
j. c * 26 * , 67 ( 2002 ) ; s.s .
agaev , phys .
d * 72 * , 074020 ( 2005 ) .
braun and i.e. filyanov , z. phys .
c * 48 * , 239 ( 1990 ) .
h. eguchi , y. koike , and k. tanaka , nucl . phys .
* b763 * , 198 ( 2007 ) . p. ball , j. high energy phys . * 01 * , 010 ( 1999 ) .
p. ball , v.m .
braun , and a. lenz , j. high energy phys . * 05 * , 004 ( 2006 ) .
keum , h - n .
li , and a.i .
sanda , phys lett .
b * 504 * , 6 ( 2001 ) ; phys .
d * 63 * , 054008 ( 2001 ) .
z.t . wei and m.z .
yang , phys .
d * 67 * , 094013 ( 2003 ) ; j.w .
chen , h. kohyama , k. ohnishi , u. raha , and y.l .
shen , phys .
b * 693 * , 102 ( 2010 ) .
[ jefferson lab collaboration ] , phys .
c * 78 * , 045203 ( 2008 ) .
et al . _ ,
d * 17 * , 1693 ( 1978 ) .
a. khodjamirian , nucl . phys .
* b605 * , 558 ( 2001 ) . k.c .
yang , phys .
d * 69 * , 054025 ( 2004 ) ; t.w .
yeh , arxiv:0802.1855 [ hep - ph ] .
arnesen , i.z .
rothstein , and i.w .
stewart , phys .
b * 647 * , 405 ( 2007 ) ; erratum - ibid .
b * 653 * , 450 ( 2007 ) . | we set up a framework for the study of the power - suppressed three - parton contribution to the pion electromagnetic form factor in the @xmath0 factorization theorem .
it is first shown that the gauge dependence proportional to parton transverse momenta from the two - parton fock state and the gauge dependence associated with the three - parton fock state cancel each other .
after verifying the gauge invariance , we derive the three - parton - to - three - parton @xmath0-dependent hard kernel at leading order of the coupling constant , and find that it leads to about 5% correction to the pion electromagnetic form factor in the whole range of experimentally accessible momentum transfer squared .
this subleading contribution is much smaller than the leading - order twist-2 , next - to - leading - order twist-2 and leading - order two - parton twist-3 ones , which have been calculated in the literature . |
since the suggestion of @xmath0 suppression as a signal for the formation of the quark - gluon plasma in relativistic heavy ion collisions @xcite , understanding the interactions of the @xmath0 with other hadrons has been an important issue as @xmath0 dissociation by hadrons could also cause the suppression of the produced @xmath0 @xcite . because the @xmath0-hadron interactions can not be directly accessed by present experiments ,
the @xmath0-hadron cross sections have been estimated through several assumptions and/or model calculations .
empirically , the @xmath0-nucleon cross sections have been estimated by using the data for @xmath0 photoproduction from the nucleon @xcite , @xmath0 photoproduction from nuclei , and @xmath0 production from the nucleon - nucleus collisions @xcite .
since those data are scattered over a wide range of energy and the estimation is model - dependent , the estimated values for the @xmath0-nucleon cross sections range from @xmath4 mb to @xmath5 mb .
( see also ref .
@xcite for a recent study on this subject . )
theoretically , these cross sections have been estimated in various ways including the perturbative qcd @xcite , qcd sum - rule approach @xcite , meson - exchange models @xcite , regge theory approach @xcite , quark models @xcite , lattice qcd @xcite , and other methods @xcite . despite of the efforts to resolve the model - dependence of the cross sections ,
the uncertainties in theoretical / model calculations for the @xmath0-hadron cross sections are not yet clarified and the predicted cross sections are model - dependent not only in the magnitude but also in the energy dependence .
recently , the @xmath0-nucleon dissociation cross sections have been calculated in a quark - interchange model by hilbert _
@xcite , and the results show a large difference from the meson - exchange model predictions of ref . @xcite . in particular , the two models predict very different values for the ratio of the cross sections , depends on the energy . here , we compare the peak values of the two cross sections . ]
@xmath6 , namely , @xmath7 in ref .
@xcite , whereas @xmath8 in the model of ref .
@xcite . in this paper , in order to understand this discrepancy , we re - examine and improve the meson - exchange model of ref . @xcite by using vector - meson dominance and charm vector - current conservation to constrain the coupling constants and form factors of this model .
we will show that these constraints can lead to @xmath9 in the meson - exchange model .
we will also discuss the role driven by the tensor coupling terms of the interactions of vector mesons .
this paper is organized as follows . in the next section ,
we discuss the coupling constants of the relevant effective lagrangians for the @xmath0-nucleon absorption processes .
the @xmath0 coupling constants are discussed in connection with vector - meson dominance , which leads to the universality of the @xmath0 coupling .
the role of the conserved charm vector - current in determining coupling constants and in constraining form factors are also explained .
section iii contains the numerical results , and the dependence of the cross sections on the coupling constants and form factors are explored .
we summarize in sec . iv .
the diagrams which contribute to the @xmath0-nucleon dissociation reactions are shown in fig . [
fig : diag ] . to evaluate these diagrams , we use the following effective lagrangians : @xmath10 where @xmath11 is the @xmath0 vector - meson field and @xmath12 is the iso - doublet @xmath12 meson field , @xmath13 and @xmath14 .
the iso - doublet @xmath15 vector - meson field is defined in the similar way . in order to determine the coupling constants , several methods
have been suggested and these include quark models using the heavy quark effective theory approach @xcite , qcd sum rules @xcite , and the su(4 ) symmetry . in this section , we discuss the coupling constants in the effective lagrangians in some detail . for the @xmath0 couplings we use vector - meson dominance ( vmd ) as in refs .
@xcite . in vmd , the photon couples to a hadron through intermediate vector mesons so that @xmath16 where the sum runs over vector meson states @xmath17 and the current - field identity gives @xmath18 with the vector meson polarization vector @xmath19 .
the parameter @xmath20 can be obtained from @xmath21 .
the most recent compilation of the data gives @xmath22 , @xmath23 , @xmath24 , @xmath25 so that @xmath26 , while the su(4 ) symmetry gives @xmath27 .
this evidently shows the aspect of the badly broken flavor su(4 ) symmetry .
application to the @xmath28 meson makes the symmetry relation worse as we have @xmath29 from the data , while the su(5 ) symmetry implies @xmath30 . for the intermediate state , we consider @xmath31 by expecting either that the higher vector meson contributions are suppressed or that there is a strong cancellation among them , especially in the charm sector @xcite . applying vmd to the @xmath12-meson iso - doublet @xmath32
, we then have the following relations among the coupled equations @xcite : @xmath33 which can be solved by using the su(2 ) symmetry relations , @xmath34 and @xmath35 .
thus , we have @xmath36 this means that the photon sees the charm quark charge through the intermediate @xmath0 vector - meson . , which gives @xmath37 .
this should be compared with @xmath38 determined from the experimental data for @xmath39 . ] by applying to the @xmath15 iso - doublet , we get the relation , @xmath40 . in the qcd sum - rule calculations of refs .
@xcite , this relation holds within @xmath41% . for the @xmath42 coupling
, we use the relation of the heavy quark mass limit @xcite , @xmath43 where @xmath44 is the ( average ) mass scale of the @xmath12/@xmath15 mesons .
this leads to the coupling constants @xmath45 the coupling constant @xmath46 can also be estimated through vmd . as we have seen before , the photon sees the charm quark charge through the @xmath0 .
if we apply this to @xmath47 , then we have @xmath48 that is , the universality of the @xmath0 coupling .
this is also closely related to the charm vector - current conservation as will be shown below . in ref .
@xcite , @xmath46 was estimated from the su(4 ) relation assuming that the @xmath0 belongs to the * 15 * -multiplet , which gives @xmath49 .
but , with this assumption , the @xmath0 contains significant light quark components and , as a result , @xmath46 is underestimated by a factor of @xmath50 compared with our estimate .
the tensor coupling constant @xmath51 can also be estimated by using vmd with the anomalous magnetic moment of @xmath52 .
the magnetic moment of @xmath52 has not been measured , but the quark model predicts @xmath53 @xcite , and this gives a predicted anomalous magnetic moment @xmath54 for @xmath52 . since the light @xmath55 , @xmath56 quarks form a spin-@xmath57 state in @xmath52 , the @xmath52 magnetic moment is solely determined by the charm quark . therefore
, vmd gives @xmath58 which leads to @xmath59 . in flavor su(4 ) , mesons are in a * 15 * -multiplet and baryons are in a * 20 * -multiplet , which correspond to the meson octet and baryon octet of su(3 ) , respectively . since @xmath60 ,
there are two couplings for the meson - baryon - baryon interactions as in the case of su(3 ) , and they can be related to the su(3 ) coupling constants @xmath12 and @xmath61 .
this gives the su(4 ) symmetry relations , @xmath62 and @xmath63 , of ref .
the empirical values of ref .
@xcite for strange hadrons then give @xmath64 these values are quite different from the qcd sum - rule predictions of ref .
@xcite @xmath65 this difference can affect the value of the ratio @xmath66 as will be discussed below . for the tensor coupling constant @xmath67
, there is no theoretical prediction for its value .
if we assume the su(4 ) relation again , we have @xcite @xmath68 however , it should be mentioned that the su(4 ) symmetry breaking effects can significantly alter the values of the coupling constants given in eqs .
( [ eq : nij ] ) and ( [ eq : nij2 ] ) . therefore , in this work , we investigate the role of the @xmath69 and @xmath70 interactions by varying their coupling constants .
the non - vanishing tensor coupling for the @xmath70 interaction also causes the four - point interaction that is shown in fig .
[ fig : diag](c ) . this term can be obtained by gauging the tensor interaction and it reads @xmath71 where @xmath72 is the gauge coupling constant .
as we shall see below , @xmath72 can be related to the universal @xmath0 coupling constant by charm vector - current conservation .
the production amplitudes can be written as @xmath73 with @xmath74 where @xmath75 is the @xmath76-channel amplitude , @xmath77 is the @xmath55-channel amplitude , @xmath78 is the @xmath76-channel amplitude including the anomalous @xmath79 interaction , and @xmath80 is from the contact term . the amplitudes can be obtained straightforwardly from the interaction lagrangians , e.g. , as in ref .
@xcite , and will not be given here .
now we impose the conservation condition of the charm vector - current to the production amplitudes , i.e. , @xmath81 and @xmath82 , where @xmath83 is the four - momentum of the @xmath0 .
the anomalous terms and the @xmath51 terms already satisfy this condition separately , and we have @xmath84 where @xmath85 is the four - momentum of the produced @xmath15 meson . thus , vector - current conservation leads to @xmath86 therefore , one can verify that the charm vector - current conservation leads to the vmd relation ( [ eq : vdm - rel ] ) for the coupling constants and fixes the gauge coupling constant @xmath72 .
because of the finite size of hadrons , it is required to include form factors in effective lagrangian approaches , which are functions of the momentum of exchanged ( or off - shell ) particles .
the form factors may be calculated from more microscopic theories @xcite , but here we employ a simple phenomenological form @xcite @xmath87 where @xmath88 and @xmath89 are the four - momentum and mass of the exchanged particle , respectively . therefore ,
when the exchanged particle is on its mass - shell , it has the correct normalization @xmath90 and , as @xmath91 , @xmath92 becomes a gaussian of @xmath93 with a width of @xmath94 . in this work
, we take the limit @xmath91 . however , employing such form factors violates the current conservation condition .
there is no unique way to restore current conservation with form factors and in this work we follow the prescription of ref .
@xcite , namely , current conservation is recovered by introducing the contact diagram of fig .
[ fig : diag](c ) , which is , in practical calculation , equivalent to replace the form factors by a universal one in the form of @xmath95[1-f(u ) ] .
\label{eq : cff}\ ] ] this form factor is also employed for @xmath96 production in the presence of the @xmath70 tensor interaction .
we first discuss our results without the tensor coupling terms , i.e. , by setting @xmath97 for a comparison with the results of ref . @xcite . shown in fig .
[ fig : xs ] are our results on the total cross sections for @xmath98 obtained with the couplings of eq .
( [ eq : nij ] ) .
the results depend on the cutoff @xmath99 , and @xmath100 gev is used for this calculation .
the dependence of our results on @xmath99 will be discussed later .
the dashed , dotted , and dot - dashed lines are the contributions from the @xmath76-channel , @xmath55-channel , and anomalous terms , respectively , and the solid lines are their sums .
there are several comments in comparing with the results of ref .
we first verify the conclusion of ref .
@xcite that the anomalous interaction terms give small contributions in both reactions .
however , there are several crucial differences .
since the vmd and current conservation condition require much larger @xmath46 coupling constant , this enhances the contribution from the @xmath55-channel diagram in both reactions . as a result ,
the @xmath101 reaction is dominated by the @xmath55-channel diagram , and the @xmath102 has comparable contributions from both the @xmath76- and @xmath55-channel diagrams .
furthermore , this makes the cross section ratio @xmath66 to be larger than @xmath103 , which is opposite to the result of ref .
in addition , our predictions on the energy dependence of the cross sections show more rapid decrease of the cross sections at larger energies than that of ref . @xcite .
although this is partly due to the gaussian form factor adopted in this model , it is the current conserved form of the form factors ( [ eq : cff ] ) that is mainly responsible for this energy dependence of the cross sections . taking into account all these effects
, we found that @xmath104 with our parameters , and the peak value of the cross sections for @xmath105 final state reaction is close to @xmath106 mb .
we also found that the tensor interactions , @xmath51 term and @xmath67 term , can give nontrivial contributions to the cross sections but play a different role . to see the role of these terms , we first present the results with @xmath107 but with @xmath108 in fig .
[ fig : xs - psi ] , while keeping the other parameters as in the case of fig .
[ fig : xs ] . in this case , the @xmath55-channel contributions are further enhanced .
furthermore , because of the large contributions from the @xmath51 term , the @xmath76-channel and @xmath55-channel diagrams interfere constructively in @xmath109 production and destructively in @xmath96 production , which is opposite to the results of fig .
[ fig : xs ] .
consequently , the @xmath110 cross sections are enhanced , but the @xmath111 cross sections are not changed so much .
this leads to the increase of the cross section ratio and we have @xmath112 , which is close to the prediction of the quark - interchange model of ref .
@xcite . to verify the role of the @xmath67 tensor term
, we use the value of eq .
( [ eq : nij2 ] ) since there is no theoretical prediction for this coupling constant .
other parameters are the same as in the case of fig .
[ fig : xs - psi ] and the results are given in fig .
[ fig : xs - dstar ] .
this shows that the effect of the @xmath67 term in @xmath113 is negligible , which is expected since it contributes to the suppressed @xmath76-channel anomalous term only .
however , this tensor interaction can change noticeably the cross sections for the @xmath114 reaction .
this is because the tensor term enters both in the @xmath76-channel and in the @xmath55-channel diagrams .
furthermore , this term requires the presence of the contact term . as a result
, the @xmath67 term enhances the @xmath96 production cross sections and leads to a smaller value of the ratio @xmath66 , and @xmath115 is observed in fig .
[ fig : xs - dstar ] .
the cross sections and their ratio @xmath66 depend on the coupling constants @xmath116 and @xmath117 which are not well understood yet . in order to see this dependence , we use the coupling constants ( [ eq : qcdsr ] ) predicted by the qcd sum - rule calculation of ref . @xcite .
this gives very different values of the cross section ratio , and instead of @xmath118 and @xmath119 , ( figs .
[ fig : xs - psi ] and [ fig : xs - dstar ] ) we have @xmath120 and @xmath121 , respectively .
this also changes the corresponding maximum values of the @xmath122 cross sections and they are @xmath123 mb and @xmath124 mb for the two cases , respectively , as shown in table [ tab1 ] . as was mentioned before
, the cross sections also depend on the cutoff parameter @xmath99 , and we have used @xmath125 gev . in order to see the dependence of our results on the form factor , we repeat the calculation for three different values of the cutoff , i.e. , @xmath126 gev , @xmath127 gev , and @xmath128 gev . shown in table [ tab1 ] are the peak values of the @xmath129 cross sections and the ratio @xmath66 .
we found that the cross sections for the @xmath130 reaction are more sensitive to the cutoff than those for the @xmath131 reaction .
thus the ratio @xmath66 decreases as the cutoff parameter @xmath99 increases .
however , we found that the @xmath67 term suppresses @xmath66 regardless of the cutoff parameter value .
[ cols="^,^,^,^,^,^ " , ]
in this work , we have reanalyzed and improved the meson - exchange model for @xmath0-nucleon reaction of ref .
@xcite . we found that vector - meson dominance and charm vector - current conservation lead to the universality of the @xmath0 meson couplings , which can drastically change the ratio @xmath66 of the cross sections of @xmath2 and @xmath132 .
it is also found that this ratio is sensitive to the relative strengths of @xmath116 and @xmath117 as well as to the tensor coupling terms in the @xmath133 and @xmath70 interactions .
we found that the vmd and vector - current conservation lead to a large value of @xmath66 .
this value can be further enhanced by the @xmath133 tensor interaction .
but the @xmath70 tensor interaction has the opposite role by decreasing @xmath66 . to match the quark - interchange model predictions of ref .
@xcite with those from our effective lagrangian approach leads us to conclude that @xmath116 must be larger than @xmath117 and @xmath134 must be small .
the first condition contradicts with the qcd sum - rule predictions of ref .
@xcite that prefers a similar strength for the two couplings . instead
, the su(4 ) symmetry relations satisfy this condition .
however , su(4 ) symmetry gives a large value for @xmath134 , and thus does not fulfill the second condition . since su(4 )
symmetry must be broken by the heavy charm quark mass , it would be interesting to see how badly the su(4 ) symmetry relations for @xmath116 , @xmath117 , and @xmath67 are broken .
therefore , more rigorous studies on these couplings are required , which will eventually help to reconcile the predictions of the quark - interchange model and of the meson - exchange model .
nevertheless , the constraints used in this work , vmd and charm vector - current conservation , are found to have a nontrivial role to fill the gap between the quark - interchange model and meson - exchange model predictions to some extent .
t. matsui and h. satz , phys .
b * 178 * , 416 ( 1986 ) .
j. ftnik , p. lichard , and j. pit , phys .
b * 207 * , 194 ( 1988 ) ; c. gerschel and j. hfner , , 253 ( 1988 ) ; s. gavin , m. gyulassy , and a. jackson , , 257 ( 1988 ) ; r. vogt , m. prakash , p. koch , and t. h. hansson , , 263 ( 1988 ) . j. hfner and b. z. kopeliovich , phys . lett .
b * 426 * , 154 ( 1998 ) ; k. redlich , h. satz , and g. m. zinovjev , eur .
j. c * 17 * , 461 ( 2000 ) .
d. kharzeev , c. loureno , m. nardi , and h. satz , z. phys .
c * 74 * , 307 ( 1997 ) .
f. arleo and v .- n .
tram , hep - ph/0612043 .
g. bhanot and m. e. peskin , nucl .
phys . * b156 * , 391 ( 1979 ) ; d. kharzeev and h. satz , phys .
b * 334 * , 155 ( 1994 ) ; y. oh , s. kim , and s. h. lee , phys . rev .
c * 65 * , 067901 ( 2002 ) ; s. h. lee and y. oh , j. phys .
g * 28 * , 1903 ( 2002 ) ; f. arleo , p .- b .
gossiaux , t. gousset , and j. aichelin , phys .
d * 65 * , 014005 ( 2002 ) ; t. song and s. h. lee , , 034002 ( 2005 ) .
d. kharzeev , h. satz , a. syamtomov , and g. zinovjev , phys . lett .
b * 389 * , 595 ( 1996 ) ; f. s. navarra , m. nielsen , r. s. marques de carvalho , and g. krein , , 87 ( 2002 ) ; f. o. dures , s. h. lee , f. s. navarra , and m. nielsen , , 97 ( 2003 ) ; f. o. dures , h. kim , s. h. lee , f. s. navarra , and m. nielsen , phys . rev .
c * 68 * , 035208 ( 2003 ) .
k. l. haglin , phys .
c * 61 * , 031902 ( 2000 ) ; a. sibirtsev , k. tsushima , and a. w. thomas , , 044906 ( 2001 ) ; k. l. haglin and c. gale , , 065201 ( 2001 ) ; a. bourque , c. gale , and k. l. haglin , , 055203 ( 2004 ) . y. oh , t. song , and s. h. lee , phys .
c * 63 * , 034901 ( 2001 ) ; y. oh , t. song , s. h. lee , and c .- y .
wong , j. korean phys .
soc . * 43 * , 1003 ( 2003 ) .
s. g. matinyan and b. mller , phys .
c * 58 * , 2994 ( 1998 ) .
z. lin and c. m. ko , phys .
c * 62 * , 034903 ( 2000 ) .
w. liu , c. m. ko , and z. w. lin , phys .
c * 65 * , 015203 ( 2002 ) .
illarionov and g. i. lykasov , in _ proceedings of international conference : i. ya .
pomeranchuk and physics at the turn of centuries _ , edited by a. v. berkov , n. b. narozhny , and l. b. okun , pp .
305316 , singapore , 2003 , world scientific , hep - ph/0305117 ; g. i. lykasov and w. cassing , nucl . phys . *
a782 * , 396c ( 2007 ) .
k. martins , d. blaschke , and e. quack , phys .
c * 51 * , 2723 ( 1995 ) ; c. y. wong , e. s. swanson , and t. barnes , , 045201 ( 2000 ) ; , 014903 ( 2002 ) , , 029901(e ) ( 2002 ) ; t. barnes , e. s. swanson , c. y. wong , and x. m. xu , , 014903 ( 2003 ) ; m. a. ivanov , j. g. krner , and p. santorelli , phys .
d * 70 * , 014005 ( 2004 ) .
j. p. hilbert , n. black , t. barnes , and e. s. swanson , nucl - th/0701087 .
k. yokokawa , s. sasaki , t. hatsuda , and a. hayashigaki , phys .
d * 74 * , 034504 ( 2006 ) . h. g. dosch , f. s. navarra , m. nielsen , and m. rueter , phys .
b * 466 * , 363 ( 1999 ) ; a. sibirtsev and m. b. voloshin , phys .
d * 71 * , 076005 ( 2005 ) .
a. deandrea , g. nardulli , and a. d. polosa , phys .
d * 68 * , 034002 ( 2003 ) .
r. d. matheus , f. s. navarra , m. nielsen , and r. rodrigues da silva , phys .
b * 541 * , 265 ( 2002 ) .
r. rodrigues da silva , r. d. matheus , f. s. navarra , and m. nielsen , braz .
phys . * 34 * , 236 ( 2004 ) .
m. e. bracco , m. chiapparini , f. s. navarra , and m. nielsen , phys .
b * 605 * , 326 ( 2005 ) .
d. b. lichtenberg , phys .
d * 15 * , 345 ( 1977 ) .
f. o. dures , f. s. navarra , and m. nielsen , phys .
b * 498 * , 169 ( 2001 ) .
v. g. j. stoks and .
a. rijken , phys .
c * 59 * , 3009 ( 1999 ) .
f. carvalho , f. o. dures , f. s. navarra , and m. nielsen , phys .
c * 72 * , 024902 ( 2005 ) .
b. c. pearce and b. k. jennings , nucl .
* a528 * , 655 ( 1991 ) .
r. m. davidson and r. workman , phys .
c * 63 * , 025210 ( 2001 ) . | we reinvestigate the @xmath0 dissociation processes induced by the reactions with nucleons , @xmath1 , in the meson - exchange model .
main constraints used in this work are vector - meson dominance and charm vector - current conservation .
we show that the cross section for @xmath2 can be larger than that for @xmath3 when these constraints are imposed .
the dependence of the cross sections on the coupling constants is analyzed in detail , and the comparison with the recent quark - interchange model predictions is also made . |
a two - level medium , i.e. a medium composed of a collection of two - level systems ( atoms , molecules , etc . ) , is the basic quantum - mechanical model in the theory of light - matter interaction .
it plays the same fundamental role as a harmonic oscillator in classical physics .
therefore , the comprehensive study of this model and its generalizations is one of the most important problems in theoretical optics .
the two - level model has allowed the description of a number of well - known effects @xcite including the effect of self - induced transparency @xcite . among the further generalizations of the model we single out the one which takes into account the near dipole - dipole interactions between the elements of the dense medium due to introduction into the equations of the so - called local - field correction @xcite . in stationary regime ,
the presence of the local field results in the remarkable effect of intrinsic optical bistability @xcite . in the regime of optical pulses , the investigation of local - field effects is linked with solitonic dynamics @xcite , optical switching @xcite , etc .
it was shown @xcite that , in the dense two - level medium , the scale of population and field changes becomes so short that one can not use the common slowly varying envelope approximation in space ( sveas ) to describe light pulse propagation .
therefore , one should use the full ( not truncated ) maxwell wave equation .
this is the case in the present paper where we consider one - dimensional problem of ultrashort ( femtosecond ) light pulse propagation in a two - level medium .
two remarks should be made here .
first , as discussed in our previous work @xcite , the local - field effects depend strongly on the duration of the pulse . for the femtosecond pulses considered in this paper ,
these effects are negligible .
however , the dynamics of the pulse can not be fully described by the usual self - induced transparency ( sit ) model .
this fact leads us to the second remark . in the dense two - level medium ,
calculations using the wave equation without sveas show the slow attenuation of sit solitons as was predicted in ref .
this and similar other effects provide the rich dynamics of transient processes which are one of the main objects of this investigation .
another aim of our research is to study in detail the complex dynamics of pulse collisions in the two - level medium .
interaction of counter - propagating solitons was previously a research issue in both the cases of two - level @xcite and three - level media @xcite . in the appropriate section
, we will return to some of this works and discuss the further progress achieved in this paper .
here , it is worth noting that we try to trace step - by - step the changes in parameters of the pulses obtained at relatively short distances due to the high density of the medium .
in addition , collisions of solitons and attendant effects were theoretically and experimentally studied in different other systems such as optical fibers @xcite , photorefractive crystals @xcite , and photonic crystals @xcite . in this paper
we consider temporal solitons , while the collisions between spatial ones were reported as well @xcite .
the structure of the paper corresponds to the gradual transition from the relatively simple situations to the more complicated ones .
section ii is devoted to the case of single pulse propagation in a dense two - level medium paying attention to the transformations of light spectrum as it moves through the medium . in section iii
we consider interaction of co - propagating solitons , especially those overtaking one another .
section iv is dedicated to the case of counter - propagating pulses colliding in the medium . in particular , we analyze the situation of the so - called symmetric collision there . finally , in section v we study the asymmetric collisions between the counter - propagating solitons .
we start with the semiclassical maxwell - bloch system for population difference @xmath0 , microscopic polarization @xmath1 , and electric field amplitude @xmath2 ( i.e. normalized rabi frequency ) @xcite , @xmath3 where @xmath4 and @xmath5 are dimensionless arguments ; @xmath6 is the normalized detuning of the field carrier ( central ) frequency @xmath7 from atomic resonance ; @xmath8 and @xmath9 are the rates of longitudinal and transverse relaxation , respectively ; @xmath10 is the wavenumber , and @xmath11 is the light speed in vacuum ; @xmath12 is the normalized lorentz frequency .
here we assume that the background dielectric permittivity of the medium is unity ( two - level atoms in vacuum ) .
note that in eq .
( [ maxdl ] ) we do not use slowly - varying envelope approximation ( svea ) which can not hold true even for thin films of the medium @xcite . in this paper we consider propagation of ultrashort pulses with gaussian shape @xmath13 where @xmath14 is the pulse duration .
amplitude of pulses is measured in the units of the characteristic rabi frequency @xmath15 , which corresponds to the so - called @xmath16-pulse @xcite . for calculations
the values @xmath17 ns and @xmath18 ns are taken , so that femtosecond pulses appear to be in the regime of coherent interaction with the resonant medium .
the spectra of pulses plotted in this paper are obtained as the absolute values of the fourier transform of the corresponding field profiles .
the spectra are normalized on the peak value of the incident pulse spectrum which is recognized as the unity .
we use the following parameters of calculations which hold true throughout the paper if the other is not stated : @xmath19 , @xmath20 s@xmath21 , @xmath22 , @xmath23 @xmath24 m , @xmath25 fs . in fig .
[ fig1 ] we demonstrate the results of propagation simulation for the pulse with high peak amplitude ( @xmath26 ) . as it was shown in ref .
@xcite , an ultrashort pulse in dense resonant medium experiences compression .
the reason for this is perhaps the effect of self - phase modulation reported in ref .
the effect of compression is characterized by a certain distance of optimal compression : after this distance the pulse slowly attenuates due to dispersion . for the layer of approximately optimal thickness ,
the main part of the pulse is transmitted and compressed , while some part of its energy is absorbed and then reemitted [ fig . [ fig1](a ) ] .
the spectrum of the main part has the typical bell shape , while the reemitted light has a dip at the resonant wavelength [ fig .
[ fig1](d ) ] .
this can be connected with effective reabsorption of low - intensity light at the resonant frequency . at the same time
the main , high - intensity part of the pulse is almost not absorbed due to the self - induced transparency ( sit ) mechanism .
the resulting spectrum of the transmitted radiation is shown in fig .
[ fig1](b ) : the two peaks are situated at both sides of the central wavelength .
the low - level reflected light has a dip in its spectrum as well [ fig . [
fig1](c ) ] .
figure [ fig1 ] was obtained in the absence of the local - field correction ( lfc ) , that is without the term @xmath27 in eq .
( [ dpdtau ] ) . to evaluate the influence of local field ,
we calculate the difference between the transmitted light spectra obtained with and without lfc [ fig . [ fig2](c ) ] . it is seen that lfc leads to the significant change in spectrum ( more than @xmath28 of the maximal value ) , while the intensity profile remains almost unchanged [ fig .
[ fig2](d ) ] in accordance with the results previously reported @xcite .
the difference between spectra is mainly due to the late , reemitted radiation which interacts with the medium for a time long enough to feel the local field .
nevertheless the overall form of the spectrum is the same as one can ascertain comparing figs . [
fig2](a ) and ( b ) .
therefore , hereinafter all the results are obtained taking into account lfc .
returning to fig .
[ fig1 ] , we should once more emphasize the role of pulse `` tail '' in formation of the fine structure of the complex spectrum of transmitted light .
note that the resulting spectrum is not a simple superposition of the two others [ see fig .
[ fig1](d ) ] because of interference term which should be taken into account here .
figure [ fig3 ] shows the spectra for the transmitted pulses with different initial peak amplitude @xmath29 .
the simplest case is observed for @xmath30 , i.e. for the @xmath16-pulse .
such a pulse is only slightly compressed and can propagate for a long distance almost without change .
its spectrum is generally coincident with that of the incident pulse [ fig . [ fig3](c ) ] .
the pulses with amplitudes larger and smaller than @xmath31 demonstrate significant pulse transformations .
low - intensity pulse ( in our case @xmath32 ) experiences strong absorption on the central frequency , so that the output radiation has no any regular profile [ it is not seen in fig . [
fig3](a ) due to its proximity to zero ] . the overall energy leaving the medium in this case is about half of the initial pulse energy ( for the calculational conditions denoted in the caption of fig .
[ fig3 ] ) .
this output energy is mainly concentrated in the side bands of the spectrum as seen in fig . [ fig3](b ) . in the case of high - intensity pulse ( @xmath33 ) , absorption is not significant and almost all energy transmits through the medium . however , its spectrum [ fig .
[ fig3](d ) ] implies that some part of the radiation is transferred from the central wavelength to the side bands .
the next step is to consider the transformations of spectra with distance traveled by the pulses in the two - level medium ( we increase the distance by the factor of @xmath34 , i.e. to @xmath35 ) .
we find that the spectrum of @xmath16-pulse gets more distorted as it propagates inside the medium for a longer distance [ fig .
[ fig4](c ) ] . in fig .
[ fig4](b ) one can see that the spectrum of low - intensity transmitted pulse has the same form as previously . in this figure
the spectrum of reflected light is plotted as well .
reflected radiation is concentrated mainly very close to the central ( resonant ) frequency . in other words ,
the medium in this case can serve as a very - narrow - bandwidth filter . however , the efficiency of such a filter is not large because only a few percent of the incident light is reflected .
the case @xmath36 is of maximal interest to us due to the strongly pronounced role of the pulse `` tail '' .
comparison of figs .
[ fig3](d ) and [ fig4](d ) shows that the side bands move away from the resonant wavelength as the pulse propagates in the medium .
this is also proved in fig . [ fig5](a ) and ( b ) .
this transformation of spectrum is due to change in the spectrum of the `` tail '' : the central dip gets wider [ fig .
[ fig5](d ) ] .
note that the main part of the pulse is characterized by the same spectrum [ fig .
[ fig5](c ) ] .
it remains almost unchanged because the main part demonstrates invariant propagation in the regime of self - induced transparency ( sit - soliton ) .
however , the main pulse very slowly attenuates though not so fast as was reported previously @xcite .
this is due to higher accuracy of our calculations in the current investigation . if the pulse amplitude is high enough ( @xmath37 ) , the splitting of the pulse into several solitons can be observed .
the examples of spectral transformations connected with this splitting are shown in fig .
comparison between the smooth spectra of separate solitons and the resulting jagged one implies that the phase relations between frequency components plays an important role .
this interference is especially characteristic for the pulse splitting into three components ( @xmath38 ) . in this case
[ fig6](d ) ] , apart from the central band , the spectrum has several pronounced and wide side bands .
it is also worth noting that the first soliton has always the widest spectrum while every next one is more narrow .
this corresponds to the fact that the first pulse is more compressed than any consecutive one .
now let us consider the process of interaction between two pulses co - propagating in the dense two - level medium .
some preliminary results were published previously @xcite . however , in this paper we present extended and more accurate calculational results .
first , in fig . [ fig7](a ) we show how the intensity of the second pulse changes under the influence of the first one .
the time interval between the peaks of two incident co - propagating pulses is taken to be @xmath39 .
it is seen that the first pulse loosing part of its energy causes some increase in the intensity of the second one in comparison with the case of the single pulse ( this corresponds to the dotted lines ) .
this is especially evident in the case of fig . [
fig7](a4 ) .
figures [ fig7](b ) demonstrate the transformations of spectra due to simultaneous compression and interactions between the pulses .
note that the spectrum of the pair @xmath40 [ fig .
[ fig7](b1 ) ] is similar to the spectrum of the single pulse with amplitude @xmath41 [ fig . [ fig4](d ) ] and can be treated as composed from the spectra of pulses with @xmath42 and @xmath31 [ fig . [
fig4](b ) and ( c ) ] .
other spectra are characterized by appearance of strong side lobes and complex fine structure in the central part as seen in fig .
[ fig7](b2 ) . one particular case ( the pair @xmath43 ) is worth considering in detail here .
this is because the second pulse of this pair propagates in the medium faster than the first one as results from fig .
[ fig4](a ) .
this means that , though the @xmath16-pulse does not influence the second pulse at short distances , their interaction should be getting stronger as they co - propagate in the medium . the reason for this distinction in pulse velocity can be easily understood on the basis of the well - known expression for the stationary pulse @xcite , @xmath44 where @xmath45 is the group velocity of the pulse , @xmath11 is the vacuum light speed , @xmath46 is the extinction coefficient , @xmath47 is the concentration of two - level atoms , @xmath14 is the stationary pulse duration . here
we are interested only in relative speed of both pulses , so one can write @xmath48 or , taking into account that the peak intensity @xmath49 , @xmath50 this condition is perfectly satisfied as seen in fig .
[ fig4](a ) ( for the case @xmath51 ) .
the concentration dependence also holds true as demonstrated by our careful examinations .
this fact can be treated as one more proof of validity of our calculational scheme .
thus , we can study the process of collision of two pulses co - propagating with different velocities .
this case is demonstrated in fig .
it is seen that the second , more intensive pulse overtakes the first one and , finally , at the length approximately @xmath52 they form a single pulse .
this is indicated by the dramatic simplification of the spectrum which retains only one , central lobe . as the distance increases , this single pulse disintegrates , and the intensive pulse leaves the other behind .
this fact corresponds to significant increase in complexity of the spectrum .
the results of fig .
[ fig8 ] also show that the intensity of the single pulse after coalescence is intermediate between the intensities of the pulses before the collision .
the change in the peak intensities of the pulses as they co - propagate in the medium is depicted in fig .
it is interesting to compare them with the behavior of intensity of a single pulse .
it is seen that , after some transient process , the peak intensity of the single pulse remains almost invariable .
this corresponds to the regime of solitonic propagation with the envelope perfectly described by hyperbolic secant function ( this means that @xmath53-pulse with initial area of @xmath54 transforms into a common @xmath16-soliton ) .
however , the intensity of the pulse is not strictly constant , it slightly fluctuates ; so we can not talk about strict , _ mathematical _ solitons but rather about realistic , _ physical _ ones .
in addition , the amplitude of these fluctuations tends to decrease with the propagation distance .
however , at larger distances the pulse slowly attenuates .
for example , the intensity of the pulse with initial amplitude @xmath53 drops from approximately @xmath55 ( at @xmath56 ) to @xmath57 ( at @xmath58 ) .
the next question is connected with the nature of the transient process : at first , the pulse is strongly compressed ( on the distance of about @xmath59 ) , and only then its envelope transforms to the solitonic one ( hyperbolic secant ) which is connected with some decrease in the pulse intensity .
the initial compression of the pulse is due to the self - phase modulation effect and is strongly dependent on the form of the incident pulse .
this compression can be observed even for the pulses with initial envelope described by hyperbolic secant .
it is known that , at the same area ( say , @xmath60 ) , the @xmath61-pulse is wider than the gaussian one , so that if the soliton is formed from the initial pulse , it would obtain greater duration .
but it is not the case : the soliton forms from the compressed pulse .
this compression can be characterized by the length value , namely , the so - called distance of optimal compression . returning to the collision of two pulses ,
we see ( fig . [ fig9 ] ) that they strongly influence one another , but , after the change of the propagation order , the distance between them increases and the solitons are restored .
one can say , that this figure demonstrates the interaction between the already formed or just forming solitons which are collision resistant .
from the previous section we know that the collision of co - propagating pulses does not prevent formation of stationary pulses with the same characteristics as in the single pulse case .
the collision of counter - propagating pulses has fundamentally different results .
the reasons were studied previously in refs .
@xcite . from mathematical point of view
, two counter - propagating pulses can not provide a stationary ( solitonic ) solution .
physically , this means that the absorption of some part of energy of the pulses occurs in the region of collision .
further we consider the dynamics of such collisions in detail .
we start from the collision of two gaussian pulses with the amplitude @xmath31 entering the medium at the same instant of time . in such symmetric geometry
the light appearing from both ends of the layer ( we can conditionally call it transmitted and reflected light ) will be described by the same envelope .
these envelopes are shown in fig .
[ fig10 ] for different thicknesses of the layer .
it is seen that the result of the collision strongly depends on the thickness @xmath62 and , as @xmath62 increases , the long precursor and tail become apparent ( in ref .
@xcite only the tail was observed ) .
the spectra of the precursor and the tail are similar ( fig .
[ fig11 ] ) : they have a dip in the region of resonant wavelength , while radiation at this central frequency is strongly absorbed .
to study this thickness dependence of the collision , we simulate the process of interaction between the pulses in a wide range of thicknesses @xmath62 .
the results of these calculations for the peak intensity of the transmitted pulse are demonstrated in fig .
[ fig12 ] .
it is seen that at relatively small length there are pronounced features which have a structure of something like resonances . for larger thicknesses the amplitudes of these resonances
diminish , so that the peak intensity of the pulse tends to the approximately constant value .
this implies that the resonances observed are connected with some transient process .
this process can be traced using distribution of population difference and light intensity inside the medium at different instants of time ( fig .
[ fig13 ] ) .
the thickness of the layer ( @xmath63 ) is large enough for @xmath16-solitons to be formed by the time of collision .
this stationary pulse is characterized by the typical population difference profile ( drop to full inversion and following rise to the ground state ) .
one can easily ascertain that , at the location of collision , a strong absorption really occurs , so that some part of light energy remains inside the medium even after the pulses have gone away from each other .
however , after collision , pulse propagation is not simple : after reaching minimum at @xmath64 , the pulse intensity grows ( @xmath65 ) , then goes down again ( @xmath66 ) , and so on .
this process is obviously connected with fluctuations depicted in fig .
[ fig9 ] and results in precursor and tail formation .
it is reasonable to suggest that the resonances of fig .
[ fig12 ] are due to pulse exiting in different moments of this transient process .
but if the thickness of the layer is large , these fluctuations become smaller and smaller , so that at the output we have almost stationary pulse with the area @xmath16 again and somewhat decreased intensity ( @xmath67 ) .
the next important feature of pulse collisions is their dependence on the duration ( or , equivalently , peak intensity ) of the initial pulses .
this dependence is shown in fig . [ fig14](a ) and is in accordance with the results of ref .
@xcite : for short and high - intensity @xmath31-pulses , the influence of collision on pulse propagation is not significant . on the contrary , long and low - intensity pulses strongly interact with each other and , if the duration is large enough , the main pulse ( soliton ) disappears leaving only a precursor and a tail [ see fig .
[ fig14](b ) for the pulses with @xmath68 fs ] .
note that the area of the main pulse ( @xmath60 ) is conserved even in the case of dramatic decrease of its peak intensity .
for example , the pulse corresponding to @xmath69 fs in fig .
[ fig14](b ) can be easily fitted by a hyperbolic secant function .
this means that such low - intensity solitons undergo substantial broadening .
in previous section we considered the process of collision of the two identical counter - propagating pulses entering the medium at the same instant of time .
such geometry of the problem can be referred to as the symmetric one .
further we discuss the case of asymmetric collisions when the incidence time or intensity is different for both initial pulses .
figure [ fig15 ] shows the results of collision of the pulse with the amplitude @xmath31 with the identical pulse which enters the medium at different time .
negative values of the incidence time difference @xmath70 means that the counter - propagating pulse is launched earlier than those propagating in the forward direction ( @xmath71 corresponds to the symmetric situation considered in the previous section ) . in the case of @xmath72 ,
the collision occurs closer to the input of the forward - propagating pulse , so that it has much time to form a stationary pulse after the collision .
therefore , the fluctuations of the peak intensity [ fig . [ fig15](a ) ] are less pronounced in this case than for @xmath73 . on both edges of this dependence ( that is , for large absolute values of @xmath70 )
one obtains the trivial case of the single pulse propagation .
the transition to this extreme regime has jump - like character on both sides of the peak intensity dependence .
this symmetry does not take place for the time of pulse transmission through the layer [ fig . [
fig15](b ) ] .
for large negative @xmath74 the abrupt increase of time propagation occurs due to formation of the low - intensity ( and , hence , slow ) soliton near the very input .
as the counter - propagating pulse enters the medium later and later , the fast initial pulse has enough time to pass a large fraction of distance before the collision happens . as a result ,
the propagation time decrease gradually as seen in fig .
[ fig15](b ) .
thus , the collisional dynamics discussed imply the possibility of controlling such pulse parameters as its peak intensity and transmission time by proper choice of incidence time of the counter - propagating pulse .
another approach to controlling the parameters of the transmitted pulse is to use the counter - propagating pulses of different intensities . in fig .
[ fig16 ] one can see the change in peak intensity of @xmath31-pulse as a function of the amplitude of the counter - propagating ( controlling ) one . as this amplitude grows , the peak intensity of the transmitted forward - propagating pulse decreases as well as its propagation velocity .
finally , in the region of amplitudes near @xmath41 , the transmitted solitonic pulse is absent _ per se_. we have only the precursor and the tail . for larger amplitudes of controlling pulse
, the transmitted one appears again , of course , with larger retardation .
however , collisional dynamics in this case are even more complex due to the effects of pulse splitting .
therefore , we will not consider the details of these multi - pulse interactions in the present investigation .
instead , we turn to the case of complete soliton disappearance at the output of the medium when the amplitude of the controlling pulse equals @xmath41 .
the dynamics of level population difference and light intensity inside the medium depicted in fig .
[ fig17 ] imply that the primary importance in this effect is likely to be connected with the residual excitation of the medium after passing the controlling pulse .
since its initial area is @xmath54 ( amplitude @xmath41 ) , before the @xmath16-soliton formes , some part of the energy is absorbed giving rise to the pronounced tail .
it is seen that the collision leads to the significant loss of energy of the forward - propagating pulse while the controlling ( high - intensity ) pulse remains almost unchanged .
moving farther , the low - intensive pulse meets the tail of the counter - propagating one and the residual excitation and , finally , undergoes ultimate absorption in the medium .
this absorbed energy gives fluorescent radiation in the long run as was stated in ref .
in this paper , we have considered the ultrashort ( femtosecond ) pulse interaction with a dense collection of two - level atoms . in other words ,
the light - medium interaction was investigated in the coherent regime when incoherent phenomenological relaxation can be neglected .
our study was conducted in semiclassical approximation , which is valid as we are not interested in the detailed description of the processes of spontaneous radiation @xcite .
in addition , semiclassical approach allows us to correctly take into account the local field correction @xcite .
our calculations were performed for two - level atoms located in vacuum which corresponds to dense gaseous media . in order to consider solid - state systems ( two - level atoms embedded in a dielectric , for example , quantum dots in semiconductor or glassy environment ) , one has to take into account the so - called local - field enhancement factor , as it was done in stationary regime @xcite . however , if the background dielectric is not absorptive , behavior of optical pulses reported in this paper is expected to be still valid for this more complex case ( with the corresponding renormalization of the electric field ) . in conclusion
, we should discuss the prospects of pulse collision dynamics studied in the paper .
our accurate calculations show the self - healing effect for the co - propagating solitons but this is not the case for the counter - propagating pulses .
together with the transient processes of soliton formation , the strong interaction of the counter - propagating pulses provides for the rich dynamics of population difference and light intensity .
the proper choice of the parameters of the colliding pulses allows one to effectively control the characteristics of light at the output of the medium .
for example , one can obtain almost entire absorption of the soliton whose energy can be stored inside the medium , at least , for the relaxation time .
the conditions of controlling pulse intensity by choosing the intensity or incidence time of the counter - propagating one can be considered as a basis of the peculiar logic gates or other elements for which the possibility of realization is still to be studied .
30 l. allen and j.h .
eberly , _ optical resonance and two - level atoms _ , ( wiley , new york , 1975 ) .
kryukov and v.s .
letokhov , sov .
* 12 * , 641 ( 1970 ) .
mccall and e.l .
hahn , phys .
rev . * 183 * , 457 ( 1969 ) .
poluektov , yu.m .
popov , and v.s .
roitberg , sov .
* 17 * , 673 ( 1975 ) .
c.m . bowden and j.p .
dowling , , 1247 ( 1993 ) .
m. e. crenshaw , , 053827 ( 2008 ) .
hopf , c.m .
bowden , and w.h .
louisell , , 2591 ( 1984 ) . y.
ben - aryeh , c.m .
bowden , and j.c .
englund , , 3917 ( 1986 ) .
r. friedberg , s.r .
hartmann , and j.t .
manassah , , 2446 ( 1989 ) .
hehlen , h.u .
gdel , q. shu , j. rai , s. rai , and s.c .
rand , , 1103 ( 1994 ) .
bowden , a. postan , and r. inguva , , 1081 ( 1991 ) .
afanasev , r.a .
vlasov , o.k .
khasanov , t.v .
smirnova , and o.m .
fedorova , , 911 ( 2002 ) .
crenshaw , m. scalora , and c.m .
bowden , , 911 ( 1992 ) .
m. scalora and c.m .
bowden , , 4048 ( 1995 ) .
novitsky , , 015802 ( 2010 ) .
novitsky , , 023828 ( 2009 ) .
afanasev , v.m .
volkov , v.m .
dritz , and b.a .
samson , , 165 ( 1990 ) .
shaw and b.w .
shore , , 1127 ( 1990 ) .
a. pusch , j.m .
hamm , and o. hess , , 023805 ( 2010 ) .
q - han park and h.j .
shin , , 4643 ( 1998 ) .
loiko and c. serrat , , 063809 ( 2006 ) .
o. cohen , r. uzdin , t. carmon , j.w .
fleischer , m. segev , and s. odoulov , , 133901 ( 2002 ) .
b.a . malomed and s. wabnitz , , 1388 ( 1991 ) . w. krlikowski and s.a .
holmstrom , , 369 ( 1997 ) .
w. krlikowski , b. luther - davies , c. denz , and t. tschudi , , 97 ( 1998 ) .
m. petrovi , m. beli , c. denz , and y.s .
kivshar , las .
photon . rev . * 5 * , 214 ( 2011 ) .
m. erkintalo , g. genty , and j.m .
dudley , opt .
express * 18 * , 13379 ( 2010 ) .
crenshaw , , 3559 ( 1996 ) .
crenshaw and c.m .
bowden , , 1139 ( 1996 ) .
novitsky , , 18 ( 2011 ) . | propagation of ultrashort optical pulses in a dense resonant medium is considered in the semiclassical limit . in our analysis , we place emphasis on several main points . first , we study transformations of spectra in the process of pulse propagation and interactions with another pulse .
the second point involves the transient processes ( including pulse compression ) connected with self - induced transparency soliton formation inside the medium .
finally , the third aspect is the study of collisions of co- and counter - propagating pulses in the medium . in the last case
, the investigation of symmetric and asymmetric collisions shows the possibility of effectively controlling the parameters of transmitted radiation . |
it is known that one of the properties of linear diffusion equation solution @xmath3 is the infinite velocity of perturbations propagation .
it is seen , for example , from the fundamental solution of this equation @xmath4 here @xmath5 is the fundamental solution presented in the gaussian form , @xmath6 is concentration , @xmath7 is the diffusion coefficient , @xmath8 defines the unit step heaviside function , @xmath9 is associated with current coordinate , @xmath10 is the current time .
this solution contradicts to intuitive speculations that diffusive particles and any information about their movement are propagated with finite velocities .
we should mark that equation ( [ eq:1 ] ) is a consequence of supposition that principles associated with locality and local thermodynamic equilibrium are valid . in the frame of this supposition
one can use the diffusion fick s law in standard form @xmath11 connecting the current particle function and the concentration gradient . by analogy , for conductive heat transfer
one can write the fourier law connecting the heat flux with temperature gradient and for liquids and gases involved in the filtration phenomenon in porous medium the darcy s law connecting the stream of liquid with pressure gradient . one of the approaches related to resolution of the problem connected to corrections for the infinite velocity of perturbations propagation lies in the usage of linear nonlocal models , for example , suggested by the extended of nonequilibrium thermodynamics @xcite , @xcite . in this case in expression ( 3 ) some additional relaxation terms are introduced . in the simplest case ,
the fick s law in the nonlocal interpretation is written as @xmath12 and equation ( [ eq:1 ] ) is reduced to the telegraph equation of the type @xmath13 the fundamental solution of the last one for @xmath14 case is expressed as @xmath15 where @xmath16 is the modified bessel s function of the first kind having zeroth order . the paradox of the infinite velocity is eliminated and the limiting velocity describing the perturbations propagation in a nonequilibrium medium becomes finite and is equaled to @xmath17 .
in the given paper the proper selection of the fundamental solutions plays a key role .
their knowledge determines the specific solutions of linear equations of mathematical physics as well , because the corresponding green s functions can be expressed in the form of linear combinations of fundamental solutions including their derivatives and integrals .
equations of type ( [ eq:1 ] ) for many real situations of macroscopic physics are good approximation . but together with this statement the consideration of the nonlocal effects is an actual problem for relatively small intervals of observation times and distances .
it is interesting to consider the evaluation of the limits of applicability of the local thermodynamic equilibrium hypothesis and the finding of approximate approaches related to determination of the current of the value studied and gradient of potential of the corresponding scalar field .
taking into account the fact that approach ( 4 ) is also approximate , it would be interesting to evaluate errors which arise in the usage of similar expressions in comparison with solution of the `` exact '' transfer equation which is remained unknown . the certain complexity in experimental testing in laboratory conditions the adequacy of equations ( [ eq:1 ] ) and ( 4 ) lies in nonlinear dependencies of the key parameters involved .
it creates some difficulties in interpretation of the results obtained .
so , in the given paper the dynamics of hypothetical linear systems modelling the transfer phenomenon is considered .
it is necessary to remind here about the inertia heat phenomenon when nonlinearity leads also to the finite velocities of perturbations @xcite but similar and relatively exotic models are out of the scope of this research .
if one come back to micro - level the diffusion phenomenon , conductive heat transfer and filtration of liquids and gases in porous media and similar phenomena can be described with the help of random behavior of microparticles . for diffusion and heat transfer phenomenon
one can introduce concepts as the characteristic length @xmath18 and time @xmath19 of free particles path or quasiparticles of phonon type ; for filtration we consider the characteristic distances and times associated with the hopping process of liquid droplets from one pore to another one .
hence , from microscopic point of view of the phenomena considered it is important to establish the correlations between phenomenological parameters of the corresponding theories similar to the extended irreversible thermodynamics with characteristic relaxation times and microparameters of the transfer process studied . in random walk model the hopping length @xmath18 and hopping time @xmath19 characterize the dynamics of the system considered on microlevel .
for various media and processes these microparameters are varied in wide limits .
the length of free path for phonons in crystals is changed in the range @xmath20 m that corresponds to the free path time located in the range @xmath21 s for frequencies of the phonon spectrum from the interval @xmath22 hz .
we should note that in paper @xcite with the usage of the molecular dynamics methods it was shown that characteristic relaxation times in nonlocal heat transfer model written in the form of telegraph equation ( 5 ) for solid argon crystal are very small ( @xmath23 s ) that comparable with times of the phonons free path in a crystal .
the characteristic microscopic hopping length in diffusion phenomenon in the most cases is located in the range @xmath24 m , for filtration
@xmath25 m and the corresponding characteristic times are located in the intervals @xmath26 s and @xmath27 s , accordingly .
the random walk model at present time is widely used for description of many processes ; in particular , the fractal dynamic models are developed in @xcite .
supposing that on validity not only nice equations can pretend but also the nice algorithms , as well , we will be based on supposition that true description of diffusion phenomenon is determined by random walk model and there is unknown linear equation which correctly describes of the transfer phenomenon of diffusion type . in this paper the simplest variant of the random walk model for @xmath14 system is considered . in this model
the hoppings of a particle ( or quasiparticles ) are realized with equal probabilities during the interval of time @xmath19 for distances @xmath28 or @xmath29 .
then for the discrete variant of the problem formulated one can obtain the solution for the probability @xmath30 of a particle location in coordinate @xmath31 in the moment @xmath32 in the form @xcite
@xmath33 it is known that expression ( 2 ) presents one of the partial cases of solution of the random walk problem describing the probability density distribution of a random value .
really , if divide ( 7 ) on @xmath34 and put @xmath35 at @xmath36 , then one can obtain expression for the probability density function describing a probability of location of diffusive particle in the vicinity of the point @xcite @xmath9 in the form ( 2 ) . in this case ,
simple calculations show that the velocity @xmath37 of perturbations propagation accepts the infinite value ( @xmath38 at @xmath36 ) but the solution itself has automodel character with respect to variable @xmath39 .
we should note that in @xcite for simplification of expression ( 7 ) containing factorials with the usage of the stirling s decomposition formula @xmath40 only the first term in the braces is kept .
it allows to introduce a criterion for correct evaluation of this approach in the form of the following inequalities [ eq : whole ] @xmath41 @xmath42 the fundamental solution for telegraph type equation ( 4 ) represents another limiting case of the random walk problem @xcite . in this case
@xmath43 at @xmath36 .
the selection of fundamental solutions as a ground for analysis of transfer processes is justified by the fact that for linear equations of mathematical physics the knowledge of a fundamental solution determines specific solutions of these equations as well , if the corresponding initial and boundary conditions are known .
it is related to the fact that the corresponding green functions can be expressed in the form of linear combinations of fundamental solutions including their derivatives and integrals .
then at transition to continuous variables it is convenient to measure a current length in the unit of elementary length @xmath18 and the current time in the units of @xmath19 .
we put also @xmath44 , @xmath45 , @xmath46 .
then the fundamental solution of equation ( [ eq:1 ] ) is written as @xmath47 and fundamental solution for ( 4 ) as @xmath48 and the probability density distribution function of a particle located in the vicinity of the coordinate @xmath9 and in the moment time @xmath10 for model ( 5 ) and considered in the given paper as the fundamental solution of unknown transfer equations is written as @xmath49 or , being presented in the form of the euler s gamma - functions it can be presented as @xmath50 here and below the symbol @xmath51 ( gauss ) indicates on equation eq .
( [ eq:1 ] ) , the symbol @xmath52 ( telegrapher equation ) on equation ( 5 ) , and symbol @xmath53 ( random walk ) on initial equation ( 7 ) .
then we will use the following hypothesis : the true solution of evolution equation of the diffusion type presents the random walk solution in the form of fundamental solution ( 14 ) but presentations ( 2 , 11 ) and ( 6 , 12 ) present approximate approaches .
solution ( 14 ) corresponds to the unknown linear evolution equation .
let us consider the evolution of equations ( 11 , 12 and 14 ) .
on figure1 we demonstrate the values of the function @xmath54 end its gradient @xmath55 for the fixed times t= 30 and 100 .
on figure2 we show the differences of the function @xmath54 ( 14 ) from the functions @xmath56 ( 11 ) and @xmath57 ( 12 ) calculated at the same values of time .
comparison of the values and for the fixed values of time t=30 and t=100 . ] the differences of solutions ( 11 ) , ( 12 ) and ( 14 ) for the values t=30 and t=100 . ] with increasing of the time the functions considered not changing their forms are decreased in their values and are stretched in the space .
the differences between the functions @xmath56 , @xmath57 , @xmath54 become small and are decreased with the time growing . taking into account the influence of the second decomposition term figuring in the stirling s decomposition expression ( 11 ) for factorials and the asymptotic behavior of the bessel function at ( @xmath58 ) @xmath59
it is easy to calculate the values of the functions considered at the point @xmath60 in different values of time at @xmath61 .
these values are equaled approximately [ eq : whole ] @xmath62 @xmath63 @xmath64 from these relationships it follows that the differences of the functions ( 16 ) are decreased with time as @xmath0 , their relative differences as @xmath1 and their absolute values as @xmath65 .
the corresponding calculations show that for other values of the current coordinate @xmath9 the differences of the functions ( 11 , 12 and 14 ) values are decreased with time as @xmath0 . incidentally , the spatial stretching of the functions themselves at @xmath66 is proportional to @xmath2 .
let us consider the gradient concentrations dynamics corresponding to solutions ( 11 ) , ( 12 ) and ( 14 ) .
for the function @xmath54 in notations used the corresponding gradient will look as @xmath67 where @xmath68 , and for the functions @xmath69 , @xmath57 , as @xmath70 and @xmath71 for all functions considered the gradient concentrations distributions at the fixed moments of time coincide with the functions that tend to zero at @xmath72 and @xmath73 and having one maximum . for equation ( 18 ) at the fixed value of time the maximum value should be observed at @xmath74 and the value of function ( 18 ) in this point is @xmath75 , the velocity of the movement of the maximum point @xmath76 and its halfwidth @xmath77 will be equaled to @xmath78 and @xmath79 , correspondingly .
the calculations show that the differences between maximum locations points of functions @xmath80 , @xmath81 @xmath82 and their relative differences are decayed as @xmath83 and the relative changes of the maximum values of the corresponding gradients @xmath84 , @xmath85 and @xmath86 for large values of @xmath10 are equaled [ eq : whole ] @xmath87 @xmath88 let us consider now the relationship between flux and concentration gradient . for the gaussian ( 2 )
the following relationship between flux and concentration gradient is valid @xmath89 for the functions @xmath90 and @xmath91 the fluxes in the points @xmath92 ( @xmath93 ) are calculated as [ eq : whole ] @xmath94 @xmath95 the calculations lead to the following dependencies of the square differences of the functions @xmath96,@xmath90 , @xmath91 , their gradients and the corresponding fluxes at @xmath97 : [ eq : whole ] @xmath98 @xmath99
@xmath100 @xmath101 @xmath102 @xmath103 @xmath104 @xmath105 @xmath106 one can notice that deviations of the values considered from each other are decreased with time faster than @xmath1 .
so , for the temporal intervals @xmath107 calculating errors in replacement of the `` accurate '' solution by the functions , at calculations of the desired fluxes will be less then @xmath108 and for the times @xmath109 will be less then @xmath110 . we should remark also that absolute values of the fluxes and their concentration functions are relatively large only for the times @xmath111 .
for the times @xmath112 the flux value is becoming less then @xmath108 at @xmath113 , for @xmath114 is less then @xmath108 at @xmath115 , for @xmath116 is less then @xmath108 at @xmath117 .
let us consider the changings of deviations from @xmath118 for the function @xmath91 .
these deviations are defined as @xmath119 it was proved that function ( 24 ) is decreased quickly with growth of the time as @xmath120 we introduce for equation ( 4 ) the corrected function @xmath121 by means of relationship @xmath122 here the index @xmath53 stresses the origin of this relationship with random walk problem .
really , relationship ( 25 ) can present in the form of ( 4 ) for the random walk problem for @xmath14 case with replacement of the coefficient @xmath7 for the parameter @xmath123 .
the analysis of ( 25 ) shows that with the increasing of the observation time @xmath124 , but for the various values of the coordinate @xmath9 the times of approaching of the function @xmath125 to the unit value @xmath126are different as it is shown on fig.3 .
the values of the corrected function f at various values of the current coordinate x. ] we want to note here that simple and good enough analytical approximation for the function @xmath125 at the values @xmath127 and @xmath66 serves the following expression @xmath128 in this expression the function @xmath125 is presented as the approximation to the heaviside function @xmath8 for each fixed value of the coordinate @xmath9 and depends only on the characteristic velocity @xmath37 associated with random walk step . from this observation
it is easy to conclude that for each moment of time @xmath129 if we define deviation of the function f from the unit value as @xmath130 the value of the corresponding coordinate corresponding to the interval @xmath66 can be written as @xmath131 .
if we put , for example , @xmath132 , @xmath97 then @xmath133 .
it signifies that the relative velocity of approaching of @xmath125 to the unit value ( @xmath134 ) is considerably higher in comparison with the propagation velocity of the fixed density distribution value proportional to @xmath2 . then equation ( 25 )
can be rewritten in the form @xmath135 where we use the notations introduced above @xmath136 , @xmath137 , @xmath138 , i.e. they are expressed by means of microparameters associated with random walk problem .
we should note also the influence of fluctuations . really , it is necessary to take into account that in derivation of equation ( [ eq:1 ] ) the locality principle is used , as well .
it imposes certain limitations of the spatial scales where the diffusion type process takes place . for the equilibrium values of concentration associated with number of particles and temperature one can use evaluations of their fluctuations in correspondence with formulas
@xmath139 the following question can be posed , what is the size of a system should be to take into account the thermodynamic fluctuations of different macroscopic parameters as particles , temperature and etc . ?
it is obvious that we can not show the accurate limits and it has a relative character .
for example , for subsystem having @xmath140 particles the relative fluctuations give the value @xmath108 .
but for small values of time and distances the thermodynamic fluctuations become very large and it is impossible to describe propagation perturbations front using only deterministic models . in other words , for the times @xmath141 and distances @xmath142 the usage of mesoscopic approaches becomes necessary . in this case
we go out from the principle locality limits and continuous medium approach .
this statement coincides with the conclusions given above associated with comparison of the functions ( [ eq:1 ] ) and ( 5 ) and `` true '' function ( 14 ) .
in this paper we consider local - equilibrium and non - equilibrium models of transfer phenomena of the diffusion type .
these phenomena are considered in the framework of the supposition that true solutions correspond to the solutions of the random walk problem ( 14 ) while the models ( 1 - 3 ) and ( 4 - 5 ) are considered as approximations to the `` true '' solution .
the comparisons given above and analysis of the differences in fundamental solutions for equations ( [ eq:1 ] ) , ( 5 ) and continuous representation of the random walk function as the density distribution function for random value ( 14 ) allow to make the following conclusions : ( a ) the differences between the values of the functions considered are decreased with time as @xmath0 ; ( b ) the propagation velocity of the fixed values of the functions considered is proportional to @xmath2 ; ( c ) on relatively large values of observation times and distances the fundamental solutions becomes very close to each other .
taking into account these conclusions one can formulate the criterion of applicability of the simplest equations as the telegraph equation ( 5 ) or heat transfer ( [ eq:1 ] ) .
their criterion of applicability can be expressed in the form of inequalities : @xmath143 , @xmath144 and @xmath145 where @xmath18 , @xmath19 , @xmath146 are defined as microparameters .
we should note that for the most problems of macrophysics that conditions are satisfied for relatively small values of the observation time @xmath147 . for
more accurate calculations of the desired relationships between fluxes and concentration gradients in the framework of random walk problem one can use relationships ( 25 - 27 ) . in these relationships
the additional parameters figuring in diffusion equations are expressed in the form of microparameters related to the random walk problem . | the fundamental solutions of diffusion equation for the local - equilibrium and nonlocal models are considered as the limiting cases of the solution of a problem related to consideration of the brownian particles random walks .
the differences between fundamental solutions were studied .
it was shown that on the period of observation time and distances exceeding the time and space associated with one step of random walk the fundamental solutions of diffusion and telegraph equations are very close to each other .
in particular , these fundamental solutions are very close to the values of the probability density distribution of diffusive particles that is obtained from the accurate solution of the random walk problem .
the difference between the probability density values for the models considered is decreased with time as @xmath0 , the relative difference as @xmath1 , and velocity of propagation of these fixed distribution densities is proportional to @xmath2 . the new modified non - local diffusion equation is suggested .
it contains only microparameters of the random walk problem . |
the phase structure of qcd at high temperature @xmath0 and/or quark chemical potential @xmath1 has attracted a great deal of interest in cosmology , compact stars and heavy - ion collisions . during the last decades , significant advances have been made in our understandings of the phase structure of hot and/or dense quark matter . at the present time
, it is widely accepted that the qcd vacuum undergoes a phase transition ( into a chirally symmetric phase or a color - superconducting phase ) at sufficiently high @xmath0 and @xmath1 .
the studies of the phase structure of qcd so far have been exclusively devoted to homogeneous quark matter in bulk .
however , the qgp phase created in relativistic heavy - ion collisions has a finite size and it is not obvious that the size is large enough to allow to apply the thermodynamic limit .
if the size of the system of interest is not large enough , we need to take account of the deviations from thermodynamic calculations .
for instance , the fluctuations of order parameter , induced by the finite - size effects , would lower the critical temperature and change the order of phase transition .
there are many works which concern with finite - size effects on hadron physics ; lattice qcd calculations fuku1,fuku2,bail , gopie , qcdsf , koma , orht , ss , the finite - size scaling analyses @xcite , plasmon @xcite , baryon number fluctuations @xcite , and so on tho , davi , mus , rischke , kim , zak1,zak2,sto , amore1,he , kogut , wong , fraga .
in particular , the finite - size effects on the qcd phase diagram are studied in @xcite and @xcite . in the former ,
however , the discussion is limited to ( 2 + 1)-dimensional systems and in the latter , the phase transition discussed is not related directly with the manifestation of chiral symmetry . in this paper , we discuss the finite - size effects to the chiral phase transition . for the purpose of illustrating our approach , we adopt the two - flavor nambu
jona - lasinio ( njl ) model as an effective model of qcd .
the finite - size effects are taken into account by using the multiple reflection expansion ( mre ) @xcite .
then , we study the size dependence of the thermodynamic potential in the mean - field approximation . in the mre approximation , the finite - size effects
are included in terms of the modified density of states .
the mre has been used to calculate the thermodynamic quantities ( such as the energy per baryon and the free energy ) of finite lumps of quark matter @xcite . as far as overall structures are concerned , the results are in good agreement with those of the mode - filling calculations with the mit bag wave functions .
we emphasize that our model calculation differs from the finite - size scaling analyses ( in which finite - size effects act as an external field ; more specifically , mass ) and the studies in a finite box in that the present model is essentially based on the mit bag wave functions .
however , it should be noted that the relevance of the mre approximation is still inconclusive when it is applied to nonperturbative calculations @xcite . this paper is organized as follows . in section 2 ,
we apply the mre approximation to the njl model and calculate the @xmath0-@xmath1 phase diagram . however , the result is unacceptable because the order of chiral phase transition is changed from second to first . in section 3 ,
we reconsider the application of the mre to the chiral phase transition .
summary and concluding remarks are given in section 4 .
we consider up and down quarks to be massless and start with the following @xmath2 symmetric njl lagrangian , @xmath3,\end{aligned}\ ] ] where @xmath4 denotes the quark field with two flavors @xmath5 and three colors @xmath6 and @xmath7 is the coupling constant .
the pauli matrices @xmath8 act in the flavor space .
the ultraviolet cutoff @xmath9 gev and the coupling constant @xmath10 are determined so as to reproduce the pion decay constant @xmath11 mev and the chiral condensate @xmath12 in the chiral limit @xcite . in the mean - field approximation ,
the thermodynamic potential per unit volume @xmath13 is given by @xmath14 \left[1 + e^{-\beta(e_k - \mu)}\right ] \right\ } , \label{eqn : tdp1}\end{aligned}\ ] ] where @xmath15 , @xmath16 and @xmath17 with @xmath18 being the dynamically generated quark mass .
the density of states @xmath19 is given by @xmath20 now we consider a quark matter confined in a given sphere that is embedded in the hadronic vacuum and do not impose the pressure balance condition between interior of the sphere and external vacuum .
we would rather fix the radius of the sphere and examine the phase diagram . to incorporate finite - size effects into the thermodynamic potential
, we replace the density of states @xmath21 with the mre density of states @xmath22 , @xmath23 , \label{eqn : mre}\end{aligned}\ ] ] where @xmath24 is the radius of the sphere and the functions , @xmath25,\end{aligned}\ ] ] correspond respectively to the surface and the curvature contributions to the fermionic density of states .
the parameter @xmath26 in eq .
( [ eqn : mre ] ) should be identified with the mass of quarks in the interior of the sphere .
it should be also noted that the precise expression of the function @xmath27 has not been derived within the mre framework .
here we used the assumption proposed in ref .
we show the momentum dependence of the mre density of states in fig .
one can see that the mre density of states is suppressed by the finite - size effects and the suppression is more pronounced at low momenta .
it can also be seen that the mre density of states takes unphysical negative values at low momenta because of the neglect of the higher order terms in @xmath28 .
thus , we shall introduce the infrared cutoff @xmath29 that is defined by the largest zero of @xmath30 , as is shown in fig .
[ fig1 ] .
thus , within the mre approximation , the thermodynamic potential including the finite - size effects is expressed as follows , @xmath31 \left[1 + e^{-\beta(e_k - \mu)}\right ] \right\}. \label{eqn : tdp2}\end{aligned}\ ] ] to obtain the phase diagram we minimize eq .
( [ eqn : tdp2 ] ) with respect to @xmath26 .
when the value of @xmath26 at the minimum is zero , the system is in the chiral symmetric phase ( qgp ) and if the value of @xmath26 is not zero , the system is in the broken symmetry phase ( hadronic ) .
we then find the phase diagram given in fig .
the outermost curve corresponds to @xmath32 ( thermodynamical limit ) . in this limit , of course our approach recovers the known results@xcite .
the order of the chiral phase transition is second order at small chemical potential and as the chemical potential increases , the order of the phase transition is changed form second to first at the tricritical point . in this curve
the parts corresponding to the second order and the first order transitions are indicated by the dotted and solid lines , respectively . to see the effect of finite size in this approach
, we show the critical temperature lines for @xmath33 @xmath34 , @xmath35 , and @xmath36 fm ( from outside to inside ) .
one can see that the critical temperatures are lowered as @xmath24 is decreased .
this behavior is consistent with the enhancement of large fluctuations induced by the finite - size effects .
however , we note that the tricritical point which exists at @xmath32 disappears for all finite values of @xmath37 at any value of chemical potential , the order of phase transition becomes the first one and the second order phase transition disappears .
this is because , as we will see below , the mre can not describe the second order phase transition .
these results obtained by the simple application of the mre to the thermodynamic potential are physically unreasonable for the following reasons .
first , the reduction of the critical temperature seems to be too strong .
second , more critical one from physical point of view , is the disappearance of the second order phase transition . in general , large fluctuations of the order parameter could increase the order of the phase transition , for instance , from first to second .
therefore , it is unnatural that the second order phase transition was converted to the first order one .
this unphysical behavior indicates the necessity of some modification of the present simple application of the mre to the njl model . to clarify the problem mentioned above ,
let us take a closer look on the self - consistency condition ( scc ) including the mre density of states .
the derivative of @xmath38 with respect to @xmath26 is given by @xmath39 \left [ 1+e^{-\beta ( e_{k}-\mu ) } \right ] \right\ } \nonumber \\ & & -\nu \int_{\lambda _ { \mbox{\scriptsize ir}}}^{\lambda _ { \mbox{\scriptsize uv}}}dk\rho _ { \mbox{\scriptsize mre}}\frac{m}{e_{k}}\left [ 1-\frac{1}{e^{\beta ( e_{k}+\mu ) } + 1}-\frac{1}{e^{\beta ( e_{k}-\mu ) } + 1}\right ] \label{eqn : mscc}\end{aligned}\]][the term proportional to @xmath40 vanishes by virtue of @xmath41 .
for the second order phase transition to exist , it is necessary that this derivative vanishes at @xmath42 however , as has been pointed out in ref .
@xcite , @xmath43 is not the solution to the scc because of @xmath44 so that @xmath45 except for @xmath46 that is , the thermodynamic potential has always positive gradient at @xmath43 , hence the thermodynamic potential ( [ eqn : tdp2 ] ) can not describe the second order phase transition . in fig . [ fig3 ] , the thermodynamic potential at @xmath47 mev is plotted as a function of @xmath26 for @xmath32 , @xmath48 and @xmath49 fm .
in the proceeding section , we examined the @xmath0-@xmath1 phase diagram , looking for the minimum of the thermodynamic potential .
this sort of the application of the mre density of states has been employed in refs . @xcite .
the second order phase transition suddenly disappears soon after taking the finite - size effect into account .
thus , the phase diagram with finite size does not converge to that with infinite size even if the radius @xmath24 is increased .
it follows that we can not take the thermodynamic limit in the mre approximation .
moreover , fluctuations due to the finite - size effects is expected to change the order of phase transition from first to second .
hence , it is unacceptable that the finite size effect is taken account of the mre approximation employed in the proceeding section . in this section ,
we reconsider the application of the mre to resolve this discrepancy .
the reason why the second order chiral phase transition disappeared in the previous section is the nonexistence of the trivial solution to the scc .
the disappearance is due to the mass dependence contained in the mre density of states which prevents the disappearance of the second order phase transition is due to the mass dependence contained in the mre density of states .
as is shown in the last of the proceeding section , the scc does not have the trivial solution because of the mass dependence .
however , there can be another scenario for the application of the mre .
according to the paper by bailin and bloch@xcite , the parameter contained in the mre density of states is not the mass of particle inside the cavities but a parameter which characterizes the penetration of the wave function of the quark matter confined in a finite volume into the hadronic environment . in the following ,
we impose specific boundary conditions on quarks and study the cases in which the mre density of states does not depend on the dynamical quark mass . in this scenario ,
the mass parameter contained in the mre density of states is replaced by the parameter @xmath50 which is generally independent of the dynamical quark mass .
then , the scc is given by @xmath51.\end{aligned}\ ] ] we do not have the derivative term of the mre density of states and the scc obviously has the trivial solution .
for instance , let us consider the dirichlet boundary condition by setting @xmath52@xcite .
then , we obtain @xmath53and the infrared cutoff is given by @xmath54 in fig .
[ fig4 ] , we show the corresponding phase diagram . the solid and dotted lines represents the critical lines of the first order and second order phase transitions , respectively .
these lines represent the critical lines of ( a ) @xmath32 , ( b ) @xmath48 , ( c ) @xmath49 and ( d ) @xmath55 fm from the out - most one , respectively . compared to the proceeding result
, the finite - size effects are suppressed .
the dashed line represents the trajectory of the tricritical point as a function of @xmath24 .
one can easily see that the temperature and the chemical potential of the tricritical point are reduced as the @xmath24 is decreased .
finally , the tricritical point disappears at @xmath56 fm and hence the region of the first order phase transition vanishes .
it should be noted that the phase diagram calculated with the mre density of states converges to that in the infinite size in the limit of @xmath57 , and is consistent with the fact of the increase of fluctuations due to the finite size effects .
thus , this approach seems to be more plausible than the proceeding approximation .
note that even the second order phase transition disappears at the radius smaller than @xmath58 fm .
this is because of the existence of the infrared cutoff and we will find the phase transition at smaller @xmath24 by taking account of the higher order terms of the mre density of state .
if this boundary condition is applicable for the real physical situation created in relativistic heavy ion collisions , then the equation of state of the matter changes with the expansion of the system . at the early stage of creation of the hot matter at the center ,
the system size is the order of @xmath59 fm and the critical temperature at @xmath60 is around @xmath61 mev , but after the system expands and reaches the hadronic freezeout stage , then the system size is much larger and the critical temperature almost recovers the thermodynamical limit .
therefore , the hydrodynamical description of such a situation requires an equation of state which reflects the above finite - size effect .
the physical environment consistent with the mit bag model is considered to be realized by setting the neumann boundary condition , @xmath62 , @xmath63 and the infrared cut off is @xmath64 in fig .
[ fig5 ] , we show the corresponding phase diagram .
the lines correspond to the critical lines for the cases of ( a ) @xmath65 , ( b ) @xmath48 , ( c ) @xmath49 , ( d ) @xmath66 and ( e ) @xmath67 fm from the out - most one , respectively .
the finite - size effects are more suppressed , because the surface term vanishes in this boundary condition , and hence the difference of the critical lines between @xmath65 and @xmath68 fm in fig .
[ fig4 ] is negligible .
the remarkable reduction of the critical temperature is observed at smaller size than @xmath69 fm .
the trajectory of the tricritical point denoted by the dashed line changes steeper than the case of the dirichlet boundary condition . in this boundary condition , the tricritical point vanishes at @xmath70 fm and the second order phase transition disappears at the radius smaller than @xmath71 fm .
contrary to the previous case , results from this boundary condition indicate that the finite size system with @xmath68 fm is large enough to take the thermodynamic limit . in this case
, it will be reasonable to apply the usual equation of state which is given by the thermodynamical limit in hydrodynamical models to describe the heavy - ion collision processes .
so far , we have discussed the behavior of the thermodynamic potential .
however , it is not easy to see how the second order phase transition is affected by the finite - size effects .
for the purpose of looking at characteristic changes of the second order transition , we calculate the specific heat defined by @xmath72 note that the specific heat shows discontinuity at the critical point of the second order phase transition .
figures 6 and 7 show the specific heat of the systems for the cases of @xmath73 and @xmath74 , respectively . here , we chose @xmath75 gev for a common quark chemical potential and examined the cases of @xmath76 fm and @xmath77 fm . in both cases , the specific heats jump at the corresponding critical temperatures .
the jump of the specific heat still exists after including the finite - size effects .
the result indicates that the second order phase transitions remain second order .
however , the gap of the specific heat is decreased by the finite - size effects .
in this paper , wee have discussed the finite - size effects on the chiral phase transition in qcd in the context of the two - flavor nambu - jona - lasinio model in the mean - field approximation . to incorporate finite - size effects into the thermodynamic potential
, we employed the multiple reflection expansion ( mre ) , where finite - size effects are included by the modification of the density of state .
we applied the mre approximation in different two ways .
one is that the mass parameter contained in the mre density of state is identified as the dynamical quark mass as is done in @xcite . in this case
it is shown that only the first order phase transition can take place for any finite size and the critical point disappears .
the other is that we identify the massparameter as the inverse of the logarithmic derivative of the wave function at the surface , following the discussion by balian and bloch@xcite .
this corresponds to the situation that our system is confined in a finite volume by an external potential and the boundary condition is fixed independently from the order parameter of the system .. then , the order of chiral phase transition changes from the second to the first as the chemical potential increases .
the temperature and the chemical potential of at the critical point become smaller as the system size decreases .
this behavior is consistent with the fact that fluctuations is enhanced by the finite - size effects .
however , quantitatively , we found that the two extreme boundary conditions lead to substantially different results .
for the dirichlet boundary condition , the finite - size effect is still large . if this is the case , the critical temperature changes from @xmath78 mev to @xmath79 mev at zero chemical potential if the size of the system expands from @xmath49 fm to @xmath35 fm .
such effects should reflect on the hydrodynamical collective observables , i.e. , the @xmath80 . on the other hand , for the neumann boundary condition ,
the finite - size effects are not so visible and the equation of state is almost the same as that obtained from the thermodynamical limit .
it should be noted that the mre calculation will lose its validity when the size of our system is very small since we have ignored the higher order terms of the mre approximation .
moreover , the finite - size effects considered in the mre approximation is only the reduction of the density of state .
however , in general , we can expect other influences due to the finite - size effects ; for example , the energy eigenvalue of particles will be also affected .
we calculated the thermodynamic potential in the mean - field approximation , although it is known that large fluctuations near phase transitions invalidate the mean - field approximation .
however , the qcd critical temperature in the mean - field approximation is qualitatively coincide with that of the lattice qcd calculations .
thus , the results obtained in this work would be still qualitatively reliable , unless the fluctuations are extraordinary enhanced by the finite - size effects . finally , we comment on the outlook for future studies . in this paper
, we studied the static behavior of chiral symmetry .
however , the finite - size effects would also affect the dynamical behaviors .
for instance , the life time of fluctuations near phase transitions increases because of the critical slowing down ( csd)@xcite . the csd is due to the large fluctuations , and hence the finite - size effects will enhance the csd . | we discuss the finite - size effects on the chiral phase transition in qcd .
we employ the nambu - jona - lasinio model and calculate the thermodynamic potential in the mean - field approximation .
finite - size effects on the thermodynamic potential are taken into account by employing the multiple reflection expansion .
the critical temperature is lowered and the order of the phase transition is changed form first to second as the size of the system of interest is reduced . |
cosmic black holes ( bh ) show a very wide range of masses : from stellar masses to hundreds of millions of solar masses . an open question has long been whether the accretion and variability processes occurring in the immediate vicinity of the event horizon are the same over this broad mass range . to test this intriguing possibility , investigators identified and studied a characteristic timescale associated with the aperiodic x - ray variability of black hole x - ray binaries ( bhxrbs ) and active galactic nuclei ( agn ; see , e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
this timescale corresponds to a break in the power spectral density ( psd ) at a certain temporal frequency where a power - law of slope @xmath2 ( with @xmath3 ) breaks to a steeper slope @xmath4 .
@xcite proposed a scaling relationship between the break frequency , the mass of the black hole and the bolometric luminosity of its accretion flow .
this extended earlier work which hypothesized that the break timescales of agn scale linearly with bh mass from the timescales observed in bhxrbs , albeit with some scatter .
@xcite found that this scatter can be explained by introducing the bolometric luminosity as a correction factor .
this lead them to conclude that agn are scaled - up galactic bhs .
however , their sample consists of only 10 agn and it is therefore desirable to test their scaling relation with newly determined agn break frequencies .
the bh in our own galactic center ( identified with the radio source sgr a * ) is an an especially interesting object , as it is the most underluminous bh accretion system observed thus far ( with a bolometric luminosity nine orders of magnitude lower than its eddington luminosity ) .
it can therefore test the @xcite relation in so far inaccessible regions of the parameter space .
the existence of a supermassive bh in the center of the milky way has been demonstrated beyond reasonable doubt by the proper motion of stars detected in the near - infrared ( nir ) waveband @xcite .
x - ray and nir emission associated with sgr a * has been observable since 2001 and 2003 , respectively @xcite , which showed that it is highly variable at both wavelengths ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
simultaneous nir and x - ray monitoring campaigns revealed that each x - ray flare is accompanied by a nir flare with zero time lag , which leads to the conclusion that the x - ray photons are being produced by compton scattering off of the relativistic electrons which radiate in the nir @xcite .
the fact that the same bunch of electrons is the source for both the nir and x - ray flux makes it possible to interpret the nir timing studies reported here in the context of the agn and bhxrb x - ray results .
in fact , since the x - ray background is very high due to the lack of resolving power of x - ray telescopes ( and thus only during the brightest flares is sgr a * actually observed ; see , e.g. , * ? ? ? * ; * ? ? ? * ) , a detailed study of the psd is more complicated in the x - rays than in the nir to the observed flux when sgr a * is at its lowest flux levels @xcite .
this low background level is negligible for the purpose of our study . ] .
recent studies of the nir properties of sgr a * showed that the variability on timescales of minutes to hours is completely described by an ( unbroken ) power - law psd with a slope of -1.6 to -2.5 ( * ? ? ? * ; * ? ? ? * see also fig .
18 in eckart et al . 2006a . ) .
in this _ letter _ , we extend the time baseline up to years by combining keck and vlt data
. we report for the first time the existence of a power - law break frequency in the nir psd of sgr a * , and we show that its deduced range is inconsistent with the @xcite scaling relation , but fits the expected value when linear scaling with mass is assumed .
the nir adaptive optics instruments at keck ii ( nirc2 ) and at the vlt ut4 ( naco ) have been used for sgr a * observations since 2003 . a time baseline of years and a sampling timescale of minutes are needed to analyze the broadband psd of sgr a * from minutes to years .
obviously , large gaps in the sampling pattern are unavoidable as the nir observations have to be carried out during the night and available telescope time limits realistic data sets to a couple of nights each year .
to cover the high - frequency to mid - frequency part of the psd ( meaning timescales of minutes to days in this context ) , we looked for keck and vlt data from consecutive nights with individual night observations that are as long as possible .
we identified the nights from 2004 july 06 08 ( all times ut ) as the most suitable .
beginning on july 06 , sgr a * was observed from 07:50 to 10:35 ut with keck ii , from 23:19 to 04:16 ( july 07 ) with the vlt , then again from 06:35 to 10:30 with keck ii , and finally from 00:53 ( july 08 ) to 06:53 with the vlt ( the keck and vlt data have been published in * ? ? ?
* ; * ? ? ?
* respectively ) .
clearly , this dense coverage is only possible by combining the keck and vlt data sets .
please note that these alternating observations were coincidental and not arranged .
as the keck data for these two nights were taken at l ( centered at @xmath5 ) and all other data were taken at k - band ( centered on @xmath6 ) ) and vlt a k@xmath7 filter ( @xmath8 ) .
we calibrated both data sets in the same way and thus the difference is negligible .
] , we adopt the finding of @xcite that the nir flux is consistent with a constant spectral index of @xmath9 ( @xmath10 ) .
thus , we scaled the flux at l down to the k - band level with this relation .
the resulting lightcurve for the 3 nights is shown in fig .
all fluxes are de - reddened using @xmath11 as it was done in @xcite .
other authors have claimed variable and/or somewhat steeper spectral indices ( e.g. * ? ? ? * ) .
however , our results presented here are not very sensitive to the exact value of @xmath12 . for the coverage of the mid - frequency to low - frequency part of the temporal spectrum ( timescales of days to years ) we averaged the data of individual nights .
this leads to a lightcurve with irregular sampling of down to one day that covers a time baseline of 4 years .
the details of the whole data set are given in table 1 .
the data reduction was standard , i.e. flat fielding , sky subtraction , and correction for bad / hot pixels .
the vlt data have been deconvolved with point - spread functions ( psf ) extracted from the individual images @xcite .
aperture photometry was done on each image and the flux was calibrated relative to sources in the field with known flux .
see @xcite for more details . for the keck data
the individual psfs have been used to fit sources in the field ; see @xcite for details .
the two photometric techniques lead to indistinguishable results as shown in @xcite .
the irregular sampling with the large gaps makes standard fourier transform techniques unsuitable for our data set .
we therefore employ the first order structure function ( sf ) , defined as @xmath13 ^ 2 >
$ ] with @xmath14 being a measurement at time @xmath15 , to look for a break in the power - law psd ( a power - law in the psd translates into a power - law in the sf albeit with a different slope , see , e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the sf for our complete data set is shown in fig . [ sf ] .
the diamonds represent the high- to mid - frequency data shown in fig .
[ lc ] , and the crosses show the mid- to low - frequency part . the overlapping region was used to normalize the low - frequency part such that a continuous sf emerges .
it shows the superposition of a constant at short lags ( @xmath16min ) which is the result of uncorrelated measurement noise , a power - law portion between lags of minutes to hours , and a plateau / shallower power - law for long time lags ( @xmath17min ) .
this plateau at low frequencies points to a break in the power - law at shorter lags . to quantify the evidence for the existence of a break frequency and its value ,
monte carlo simulations are required , as the sampling pattern severely distorts any intrinsic power - law psd , and introduces the effects of red - noise leak and aliasing ( e.g. , * ? ? ?
we used the following approach for our mc simulations , which is very similar to the psresp method by @xcite and @xcite : 1 . an intrinsic ( broken or unbroken ) power - law psd model
is assumed and corresponding lightcurves are generated using the algorithm by @xcite .
this is done independently for the high - frequency and low - frequency part .
these lightcurves , which are significantly longer and more densely sampled to account for aliasing and leakage , are subsequently resampled according to the sampling function of the real data and uncorrelated measurement noise is added ( see * ? ? ?
* for more details ) .
the sf from the simulated lightcurves are calculated in the same way as is done for the data ( the renormalization is done for each pair of long and short - term simulations individually ) .
an observed @xmath18 value is determined by using the observed sf , the mean simulated sf , and error bars equal to the rms spread of the individual simulated sfs at each time lag @xcite .
the goodness of fit is computed by modeling the @xmath18 distribution for each assumed psd model , i.e. several thousand @xmath18 values are calculated using the individual realizations of the simulations instead of the observed data . the probability that the model psd can be rejected
is then given by the percentile of the simulated @xmath18 values exceeded by the value of observed @xmath18 .
the steps above are repeated to scan a range of break frequencies and power - law slopes . for each model psd
we used 100 simulations for each the high- and low - frequency part . by combining both sets we arrived at 10,000 simulations to determine the @xmath18-distribution
. the class of models we employed as the intrinsic psd is the singly broken power - law : @xmath19 here , @xmath20 is the break frequency ( the characteristic timescale ) , @xmath21 the psd amplitude at @xmath20 , @xmath22 the high - frequency power - law slope , and @xmath2 is the low - frequency power - law slope with the constraint @xmath23 .
we tested a @xmath22 range of @xmath24 in increments of 0.1 , a @xmath2 range of @xmath25 in increments of 0.1 , and a break frequency range of @xmath26 in multiplicative steps of 1.2 .
we also tested unbroken power - law models ( @xmath27 ) which are , however , rejected with a likelihood of acceptance of @xmath28 ( i.e. none of the 10,000 simulations equalled or exceeded the observed @xmath18 ) .
the best fit result with a likelihood of acceptance of @xmath29 can be found at @xmath30 ( @xmath31min ) , @xmath32 , and @xmath33 ( errors here are 90% confidence limits ) .
this solution is plotted over the observed sf in fig . [ sf ] . in fig .
[ contours ] we show the confidence contours in the @xmath34 plane at the best fit value for @xmath35 .
when the three dimensional distribution is marginalized over two parameters , we arrive at the following likeliest values and their 90% confidence limits for the individual parameters : @xmath36 ( @xmath37min ) , @xmath38 , and @xmath39 .
the existence of a break in the power - law psd of sgr a * can also be inferred by a simple argument : table 1 reveals that the mean flux of sgr a * during one night stays roughly the same over the 4 years of observations .
if the power - law psd extended unbroken over decades , one would expect to observe very different mean fluxes over a time baseline of years .
our detailed mc simulations serve the purpose of quantifying the location of the power - law break .
the same simple argument also holds true for the mean x - ray flux of sgr a*. our result of a break frequency at @xmath40 is strikingly different from the prediction of the scaling relation proposed by @xcite . using a mass of @xmath41 @xcite and a bolometric luminosity of @xmath42 erg / s @xcite ,
the relation by @xcite predicts a break frequency at the order of @xmath43 for sgr a*. interestingly , if the term with the bolometric luminosity is neglected in the @xcite relation , i.e. the bolometric index ( called b in their paper ) is set to zero , their relation predicts a break timescale of @xmath44min for sgr a*. also , our deduced break timescale fits the expected value if linear scaling with bh mass is assumed from the typical break timescales observed in the high / soft and low / hard states of cyg x-1 ; see fig .
this suggests that , while the inclusion of the bolometric luminosity as a free parameter improves the fit in a limited luminosity range , it is in fact not the true physical correction factor that can explain the scatter around the linear mass scaling .
it is , however , uncertain how meaningful a comparison of sgr a * s break timescale with the breaks of the much higher luminosity agn used to derive the mass - luminosity - timescale relation of @xcite is .
these seyferts are almost certainly in a different accretion state .
they have a prominent optically thick accretion disk and much higher accretion rates .
the @xcite relation seems to work well when extrapolated to soft state bhxrbs , which are at comparable eddington accretion rates to the seyferts , but for the low - accretion rate hard states the comparison is less obvious .
in fact , work by @xcite suggests that in the hard state the high - frequency shape of the psd is constant despite large changes in luminosity .
therefore , we conclude that the break in the nir psd of sgr a * implies that the @xcite relation with luminosity scaling applies to soft state agn , but that in the hard state there is only mass scaling .
then sgr a * seems to fit in well with the hard state bhxrbs , consistent with its low luminosity .
future work will further elucidate the relationship between variability processes across the range of bh masses .
our work presented here extends the set of supermassive bhs with determined break frequencies .
the bh in our galactic center is especially important , as it is the most underluminous source ( in terms of eddington luminosity ) accessible to observations , and its mass determination is the most precise for supermassive bhs @xcite .
we want to thank phil uttley and matt malkan for their constructive comments on the manuscript .
lm is supported by a fellowship within the postdoc - program of the german academic exchange service ( daad ) .
some of the data presented here were obtained from mauna kea observatories .
we are grateful to the hawaiian people for permitting us to study the universe from this sacred summit .
this work was supported by nsf grant ast-0406816 .
cccccc 2004 july 06 & keck & l & 165 & 2.64 & hornstein 2007 + 2004 july 07 & keck / vlt & l/k & 532 & 4.05 & hornstein 2007/eckart+ 2006a + 2004 july 08 & vlt & k & 360 & 2.00 & eckart+ 2006a + 2004 july 26 & keck & k & 42 & 4.43 & ghez+ 2008 + 2005 july 31 & vlt / keck & k & 591 & 4.01 & meyer+ 2008 + 2006
may 03 & keck & k & 140 & 5.53 & do+ 2008 + 2006 june 20 & keck & k & 125 & 4.57 & do+ 2008 + 2006 june 21 & keck & k & 164 & 3.62 & do+ 2008 + 2006 july 17 & keck & k & 189 & 2.86 & do+ 2008 + 2007 may 18 & keck & k & 84 & 5.53 & do+ 2008 + 2007 august 12 & keck & k & 57 & 3.05 & do+ 2008 + 2008 may 15 & keck & k & 153 & 4.57 & unpublished + 2008 june 02 & keck & k & 160 & 4.00 & unpublished + 2008 july 24 & keck & k & 179 & 3.51 & unpublished , crosses show the mid- to low - frequency data ( see table [ tbl1 ] ) .
the solid line is our best fit with @xmath45 , @xmath35 , and @xmath46 , see section 4 .
please note that these parameters describe the psd and can not simply be read of the sf depicted here .
error bars are determined from our monte carlo simulations as described in section 3 . ] | proposed scaling relations of a characteristic timescale in the x - ray power spectral density of galactic and supermassive black holes have been used to argue that the accretion process is the same for small and large black holes . here , we report on the discovery of this timescale in the near - infrared radiation of sgr a * , the @xmath0 black hole at the center of our galaxy , which is the most extreme sub - eddington source accessible to observations .
previous simultaneous monitoring campaigns established a correspondence between the x - ray and near - infrared regime and thus the variability timescales are likely identical for the two wavelengths .
we combined keck and vlt data sets to achieve the necessary dense temporal coverage , and a time baseline of four years allows for a broad temporal frequency range .
comparison with monte carlo simulations is used to account for the irregular sampling .
we find a timescale at @xmath1min ( errors mark the 90% confidence limits ) which is inconsistent with a recently proposed scaling relation that uses bolometric luminosity and black hole mass as parameters .
however , our result fits the expected value if only linear scaling with black hole mass is assumed .
we suggest that the luminosity - mass - timescale relation applies only to black hole systems in the soft state . in the hard state , which is characterized by lower luminosities and accretion rates ,
there is just linear mass scaling , linking sgr a * to hard state stellar mass black holes . |
top quark is the heaviest elementary particle with a mass of @xmath1 gev @xcite .
top s large mass is a reason to rapid decay so that it has no time to hadronize and if it were not for the confinement of color , the top quark could be considered as a free particle .
this property allows the top quark to behave like a real particle and one can safely describe its decay in perturbative theory .
the large hadron collider ( lhc ) is a superlative top factory with 90 million @xmath2-pairs per year which will allow us to perform accurate studies of the top - quark properties , such as its mass @xmath3 , total decay width @xmath4 , the effective weak mixing angle , and elements @xmath5 of the cabibbo - kobayashi - maskawa ( ckm ) @xcite quark mixing matrix .
the theoretical aspects of top - quark physics at the lhc are summarized in a recent paper @xcite . since @xmath6 ,
top quarks almost exclusively decay to bottom quarks , via @xmath7 within the standard model ( sm ) theory and in beyond - the - sm theories with an extended higgs sector , top quarks decay via @xmath8 .
+ many extensions of the standard model contain more than one higgs doublet , and the new degrees of freedom appear as extra higgs scalars .
for example , in the supersymmetric sm , at least two higgs doublets are needed so as to cancel gauge anomalies and to generate masses for both up- and down - type quarks @xcite ; in the weinberg model of cp violation , at least three higgs doublets are needed in order to have spontaneous cp violation(see @xcite and references therein ) . both neutral and charged
physical higgs bosons exist in all these extensions of the sm .
the observation of charged higgs bosons , @xmath9 , would indicate physics beyond the sm .
for the sake of simplicity , we will limit ourselves throughout this paper to the decay @xmath8 in a model with two - higgs - doublet , in which case there is only one physical charged higgs boson remaining after spontaneous symmetry breaking .
new results of a search for the charged higgs bosons in proton - proton collision at a center - of - mass energy of @xmath10 tev are reported by the atlas collaboration @xcite where the @xmath11jets channel in @xmath2 decays is used with a hadronically decaying @xmath12 lepton in the final state .
the reported data leads to a limit on the product of branching ratios @xmath13 of 0.03 - 0.10 for @xmath9 masses in the range @xmath14 gev@xmath15 gev . + as it is mentioned in ref .
@xcite , a clear separation between the @xmath7 and @xmath8 can be achieved in both the @xmath16 pair production and the @xmath17 single top production at the lhc .
the qcd corrections to the decay rate @xmath18 are known at next - to - next - to - leading order @xcite and the nlo electroweak corrections were found in ref .
@xcite . in this paper
, we evaluate the first order qcd corrections to the decay of an unpolarized top quark into a charged higgs boson .
since bottom quarks hadronize , via @xmath19 , therefore the decay process @xmath20 is of prime importance and the particular purpose of this paper is to make predictions for the energy spectrum of b - flavored mesons in top decay .
this measurement will be important for future tests of the higgs coupling in the minimal supersymmetric standard model ( mssm ) at lhc .
the hadronization of the bottom quark was considered in the nlo qcd analyses of the decay @xmath21 in refs .
@xcite . as it is shown in @xcite
, the hadronization of the bottom quark is identified to be the largest source of uncertainty in measurement of the top - quark mass . + to study the distributions in the @xmath22-meson scaled - energy @xmath23 , we employ the massless scheme or zero - mass variable - flavor - number ( zm - vfn ) scheme @xcite in the top - quark rest frame . in this scheme ,
the zero mass parton approximation is also applied to the bottom quark and the non - zero value of the b - quark mass only enter through the initial condition of the nonperturbative fragmentation function(non - ff ) .
nonperturbative ff is describing the hadronization process @xmath24 and is subject to dokshitzer - gribov - lipatov - alteralli - parisi(dglap ) @xcite evolution and it is scale dependent .
this paper is organized as follows . in sec .
[ sec : one ] , we study the inclusive production of a meson from top - quark decay considering the factorization theorem and dglap equations . in sec . [
sec : two ] , we present the calculation of the @xmath25 qcd corrections to the tree - level rate of @xmath26 .
we work in zm - vfn scheme neglecting the b - quark mass in our qcd corrections but will retain the arbitrary value of @xmath27 . in sec .
[ sec : four ] , we present our numerical analysis . in sec .
[ sec : five ] , our conclusions are summarized .
we study the inclusive production of a b - meson from top - quark decay @xmath28 providing that the top - quark mass @xmath3 , bottom - quark mass @xmath29 and the charged higgs boson mass @xmath30 satisfy @xmath31 .
the gluon in eq .
( [ pros ] ) contributes to the real radiation at nlo and both the @xmath32 quark and the gluon may hadronize to the @xmath22 meson . in the equation above
, @xmath33 stands for the unobserved final state .
if we denote the four - momenta of top - quark , b quark , gluon and b meson by @xmath34 and @xmath35 , respectively , therefore in the top - quark rest frame the @xmath32 quark , gluon , and @xmath22 meson have energies @xmath36 , where @xmath37 and @xmath38 .
it is convenient to introduce the scaled energy fractions @xmath39 ( @xmath40 ) .
+ we wish to calculate the partial decay width of process ( [ pros ] ) differential in @xmath23 , @xmath41 , at nlo in the zm - vfn scheme . according to the factorization theorem of the qcd - improved parton model @xcite , the energy distribution of a hadron b
can be expressed as the convolution of the parton - level spectrum with the nonperturbative fragmentation function @xmath42 , describing the hadronization @xmath43 , @xmath44 where @xmath45 is the parton - level differential width of the process @xmath46 , with @xmath33 comprising the @xmath47 boson and any other parton . here , @xmath48 and @xmath49 are the factorization and the renormalization scales , respectively . at nlo ,
the scale @xmath49 is associated with the renormalization of the strong coupling constant . in principle
, one can use two different values for the factorization and renormalization scales ; however , a choice often made consists of setting @xmath50 and we shall adopt this convention for most of the result which we shall show .
+ in next section , we present our analytic results for @xmath51 at nlo in zm - vfns .
the coupling of the charged higgs boson to the top and bottom quark in the minimal supersymmetric standard model(mssm ) can either be expressed as a superposition of scalar and pseudoscaler coupling factors or as a superposition of right- and left - chiral coupling factors @xcite . adopting the first approach
, the born term amplitude for the process @xmath52 can be parametrized as _ @xmath53 _ , and the second scheme leads to the born amplitude _ @xmath54 _ where @xmath55 and @xmath56 . in a model with two higgs doublets and generic coupling
to all the quarks , it is difficult to avoid flavor - changing neutral currents .
we , therefore , limit ourselves to models that naturally stop these problems by restricting the higgs coupling . as it is explained in ref .
@xcite , the first possibility is to have the doublet @xmath57 coupling to all bosons and the doublet @xmath58 coupling to all the quarks ( model @xmath59 ) .
this leads to the coupling factors @xmath60 the second possibility is to have the @xmath58 couple to the right - chiral up - type quarks ( @xmath61 ) , and the @xmath57 couple to the right - chiral down - type quarks ( model @xmath62 ) .
this model leads to the coupling factors @xmath63 in equations above , @xmath64 is the ratio of the vacuum expectation values of the two electrically neutral components of the two higgs doublets and the weak coupling factor @xmath65 is related to the fermi s constant @xmath66 by @xmath67 . + the total decay width of @xmath68 at lo is given by @xmath69 where @xmath70 is the klln function . in the limit of vanishing b - quark mass , @xmath71 in model @xmath59
, therefore the tree level decay width simplifies to @xmath72 for model @xmath62 one has @xmath73 when the left - chiral coupling term , proportional to @xmath74 , can become comparable to the right - chiral coupling term @xmath75 , one can not therefore naively set @xmath76 in all expressions .
for example , if we take @xmath77 gev , @xmath78 gev , @xmath79 gev and @xmath80 thus the second term in the curly brackets can become as large as @xmath81 in model @xmath62 . in this paper
we adopt , with good approximation , the born term presented in eq .
( [ gamma ] ) in both models when @xmath82 , more detail can be found in ref .
@xcite . in the following ,
we discuss the calculation of the @xmath25 qcd corrections to the tree - level decay rate of @xmath52 and we present the parton - level expressions for @xmath83 at nlo in the zm - vfn scheme . in the zm - vfn scheme , where @xmath76 is put from the beginning
, all singularities including the soft- and collinear gluon emission are regularized by dimensional regularization in @xmath84 space - time dimensions to become single poles in @xmath85 , which are subtracted at factorization scale @xmath48 and absorbed into the bare ffs according to the modified minimal - subtraction scheme ( @xmath86 ) . in this scheme
, @xmath29 only sets the initial scale @xmath87 of the dglap evolution . + adopting the on - shell mass - renormalization scheme , the virtual one - loop corrections to
the @xmath88-vertex includes both ir- and uv - singularities .
therefore , the contribution of virtual corrections into the differential decay width normalized to the born width , reads @xmath89 where , @xmath90 , the scaled mass @xmath91 is @xmath92 and _ @xmath93_. following ref .
@xcite , the renormalized amplitude of the virtual corrections can be written as @xmath94 where @xmath95 stands for the counter term and @xmath96 arises from the one - loop vertex correction .
since we neglect the b quark mass , thus the counter term of the vertex consists of the top - quark mass renormalization and the wave function renormalizations as @xmath97 where , the mass renormalization constant of the top quark reads @xmath98 and from ref .
@xcite , for the wave function renormalization constants we have @xmath99 in eqs .
( [ mass ] ) and ( [ wave ] ) , @xmath100 and @xmath101 represent infra - red(ir ) and ultra - violet(uv ) singularities which arise from the collinear- and the soft - gluon singularities , respectively .
therefore , the real part of the one - loop vertex corrections @xmath96 is given by @xmath102 where , @xmath103 , @xmath104 for @xmath105 quark colors , and @xmath106 is the spence function . +
all uv - divergences are canceled after summing all virtual corrections up but the ir - singularities are remaining which are now shown by @xmath85 .
the virtual corrections to the differential decay rate is then given by @xmath107 as it is explained in ref .
@xcite , to cancel the ir - singularities of the virtual corrections , one needs to include the real gluon emission , namely , @xmath108 . to calculate the contribution of the real corrections , we start form the definition of decay rate . as before , to regulate the ir - divergences we work in @xmath84 dimensions , therefore from the definition of decay rate
, one has @xmath109 where , the phase space element reads @xmath110 for simplicity , we choose the top - quark rest - frame and to calculate the differential rate @xmath111 normalized to the born width , we fix the momentum of b - quark in eq .
( [ decaydef ] ) . to get the correct finite term in the normalized differential decay rate , the born width @xmath112 will have to be evaluated in the dimensional regularization at @xmath113 , i.e. @xmath114 .
thus , the contribution of the real gluon emission into the normalized differential decay width is then given by @xmath115\nonumber\\ & & + \frac{1+x_b^2}{(1-x_b)_+}\big[-\frac{1}{\epsilon}+2\ln x_b+\frac{7x_b^2 - 8 x_b+7}{2(1+x_b^2)}\nonumber\\ & & + f\big]+2(1+x_b^2)\bigg(\frac{\ln(1-x_b)}{1-x_b}\bigg)_+\big\},\end{aligned}\ ] ] where the plus distributions are defined as usual .
the nlo expression for @xmath116 is obtained by summing the born term , the virtual one - loop and the real contributions .
+ since , the b meson can be also produced from the fragmentation of the emitted real gluon , therefore , we also need the differential decay rate @xmath117 in the zm - vfn scheme . to calculate the @xmath117 , as before , we start form eq .
( [ decaydef ] ) by fixing the momentum of gluon , therefore , there will be no soft singularity .
+ our results list here @xmath118\big(-\frac{1}{\epsilon}+\gamma_e-\ln 4\pi\big)+\hat a_1(x_b)\big\}\bigg\},\nonumber\\ \frac{d\hat\gamma_g}{dx_g}&=&\hat\gamma_0\frac{\alpha_s(\mu_r)}{2 \pi}c_f\bigg\{\frac{1+(1-x_g)^2}{x_g}\times\nonumber\\ & & \hspace{2cm}\big(-\frac{1}{\epsilon}+\gamma_e-\ln 4\pi\big)+\hat b_1(x_g)\bigg\},\nonumber\\\end{aligned}\ ] ]
where @xmath119 and @xmath120 are free of singularities and , in the following , their functional form will be shown .
+ to subtract the collinear singularities remaining in eq .
( [ first ] ) , we apply the modified minimal subtraction ( @xmath86 ) scheme where the collinear singularities are absorbed into the bare ffs .
this renormalizes the ffs and generates the finite terms of the form @xmath121 in the differential decay rates . + according to this scheme , in order to get the @xmath86 coefficient functions we shall have to subtract from eq .
( [ first ] ) the @xmath122 term multiplying the characteristic @xmath86 constant @xmath123 , therefore , we obtain @xmath124\big\},\nonumber\\ \frac{d\hat\gamma_g^{\overline{ms}}}{dx_g}&=&\hat\gamma_0\big\{\frac{\alpha_s(\mu_r)}{2 \pi}\big[p_{gq}^{(0)}(x_g)\ln\frac{m_t^2}{\mu_f^2}+c_f \hat b(x_g)\big]\big\},\nonumber\\\end{aligned}\ ] ] where @xmath125 are the time - like splitting functions at leading order @xcite @xmath126 and the coefficient functions @xmath127 and @xmath128 read @xmath129\nonumber\\ & & + \frac{1+x_b^2}{(1-x_b)_+}\big[\ln(x_b^2(1-y)^2)+\frac{x_b^2 - 4x_b+1}{1+x_b^2}\big]\nonumber\\ & & + 2(1+x_b^2)\bigg(\frac{\ln(1-x_b)}{1-x_b}\bigg)_+,\nonumber\\ \hat b(x_g)&=&\frac{1+(1-x_g)^2}{x_g}\big[\ln(x_g^2(1-y)^2(1-x_g)^2)-\frac{5}{2}\nonumber\\ & & -\ln(1-(1-y)x_g)\big]+\frac{1}{2(1-(1-y)x_g)^2}\big[\nonumber\\ & & 7x_g+\frac{6}{x_g}-6-(1-y)(10x_g^2 - 8x_g+12)+\nonumber\\ & & x_g(1-y)^2(4x_g^2 - 4x_g+7)\big].\nonumber\\\end{aligned}\ ] ] in this work we identify @xmath130 , so that in eq .
( [ final00 ] ) the terms proportional to @xmath131 vanish .
+ integrating @xmath116 of eq .
( [ final00 ] ) over @xmath132 , we obtain the renormalized decay rate @xmath133\big\}.\nonumber\\\end{aligned}\ ] ] this result is in agreement with refs .
@xcite and the corrected version of @xcite . as it is seen from eq .
( [ rate ] ) , in the limit @xmath134 the total decay rate is finite and proportional to @xmath135 . at the opposite limit where @xmath136 , due to the @xmath137 singularity setting the bottom - quark mass to be zero is no longer a valid approximation for the differential decay rates and our results
must be improved considering a massive b - quark .
we are now in a position to present our phenomenological results by performing a numerical analysis . in the mssm
, the mass of the charged higgs is strongly correlated with the other higgs boson masses .
the charged higgs boson mass in the mssm is restricted at tree - level by @xmath138 , ref .
this restriction does not hold for some regions of parameter space after including radiative corrections . therefore , two key phenomenological parameters in the charged higgs searches are the higgs mass @xmath27 and @xmath139 , which are model - dependent
. searches of the charged higsses over a good part of the @xmath140 plane in the mssm is a program that still has to be carried out and this belongs to the lhc experiments .
we adopt from ref .
@xcite the present limit @xmath141 gev obtained from lep .
we also adopt from ref . @xcite the input parameter values @xmath142
gev@xmath143 , @xmath144 gev , @xmath145 gev , and @xmath146 gev .
we evaluate @xmath147 at nlo in the @xmath148 scheme using eq .
( 8) of ref .
@xcite , with @xmath149 active quark flavors and the typical qcd scale @xmath150 mev . as mentioned before
, the b - quark mass only enter through the initial condition of the nonperturbative fragmentation function .
we employ the nonperturbative @xmath22-hadron ffs that were determined at nlo in the zm - vfn scheme through a joint fit @xcite to @xmath0-annihilation data taken by aleph @xcite and opal @xcite at cern lep1 and by sld @xcite at slac slc .
specifically , the power ansatz @xmath151 was used as the initial condition for the @xmath152 ff at @xmath153 gev , while the gluon and light - quark ffs were generated via the dglap evolution .
the fit yielded @xmath154 , @xmath155 , and @xmath156 .
to study the scaled - energy ( @xmath23 ) distribution of the bottom - flavored hadrons produced in top - quark decay , we consider the quantity @xmath157 . in fig .
[ fig1 ] , we show our prediction for the size of the nlo corrections , by comparing the lo ( dotted line ) and nlo ( solid line ) results , and the relative importance of the @xmath152 ( dashed line ) and @xmath158 ( dot - dashed line ) fragmentation channels at nlo , taking @xmath159 and @xmath79 gev . the same nlo ffs are used for the lo result .
[ fig1 ] shows that the nlo corrections lead to a significant enhancement of the decay rate in the peak region and above .
furthermore , the peak position is shifted towards higher values of @xmath23 .
the gluon fragmentation leads to an appreciable reduction in decay rate at low-@xmath23 region , for @xmath160 .
for example , the gluon fragmentation decreases the size of decay rate up to @xmath161 at @xmath162 . for higher values of @xmath23 ,
the @xmath152 contribution is dominant . as we explained in section
[ sec : one ] , the mass of b - meson is responsible for the appearance of the threshold at @xmath163 .
@xmath157 as a function of @xmath23 in the zm - vfn ( @xmath76 ) scheme with @xmath79 gev and @xmath159 .
the nlo result ( solid line ) is compared to the lo one ( dotted line ) and broken up into the contributions due to @xmath152 ( dashed line ) and @xmath158 ( dot - dashed line ) fragmentation . ] in fig .
( 2b ) of ref .
@xcite , the unpolarized rate is plotted as a function of @xmath139 for a sample value of @xmath79 gev .
it can be seen that the rate is quite small for @xmath139 values exceeding @xmath164 .
+ in fig .
[ fig2 ] , we study the energy distribution of the b - meson in different values of the @xmath139 , i.e. @xmath165 and 15 , for @xmath79 gev . it can be seen that when @xmath139 is increased the decay rate is decreased , as @xmath166 is proportional to @xmath167 .
@xmath157 as a function of @xmath23 in different values of @xmath168 and @xmath169 , with @xmath79 gev .
when the values of @xmath139 exceed @xmath164 , the decay rate becomes quite small @xcite . ] in fig .
( 2a ) of ref .
@xcite , the unpolarized rate is also shown as a function of @xmath170 for @xmath159 .
the functional behavior of the rate shows that the rate is largest when @xmath171 and it drops to zero when @xmath172 . @xmath157 as a function of @xmath23 in the zm - vfn scheme with different values of @xmath173 and 150 gev ( @xmath159 ) .
thresholds at @xmath23 are shown . ]
@xmath23 spectrum in top decay considering the decay modes @xmath174 ( solid line ) and @xmath175 ( dashed line ) , taking @xmath176 gev , @xmath79 gev and @xmath177 . ] adopting the limit @xmath141 gev from ref .
@xcite , in fig .
[ fig3 ] we study the energy distribution of the b - meson in different values of the higgs boson mass , i.e. @xmath173 and 150 gev , by fixing @xmath159 .
as mentioned , the mass of b - meson creates the thresholds at @xmath178 ( for @xmath179 gev ) , @xmath180 ( for @xmath181 gev ) , @xmath162 ( for @xmath79 gev ) and @xmath182 ( for @xmath183 gev ) . in comparison with the born rate @xmath184 gev ,
the rate into a charged higgs is generally quite small except for small @xmath139 values .
one finds equality of the rates into a @xmath185 and @xmath47 only at @xmath186 for @xmath79 gev .
however , such a small @xmath139 value is excluded by the indirect limits in the @xmath187 plane @xcite . in fig .
[ fig4 ] , we compared the energy distribution of the b - meson produced in decay modes @xmath174 and @xmath175 with @xmath177 , @xmath79 gev , @xmath188 gev and the parton - level differential rates @xmath189 and @xmath190 extracted from ref .
the thresholds appear at @xmath178 ( for @xmath174 ) and @xmath162 ( for @xmath175 ) .
+ the total top - quark decay width is obtained by summing the two partial widths @xmath191 and @xmath192 order by order .
however , as figs .
( [ fig2 ] ) and ( [ fig3 ] ) show the radiative corrections coming from the supersymmetric sector depend extremely on the higgs mass and @xmath139 , but fig .
[ fig4 ] shows that the contribution of the decay mode @xmath193 is always larger than the one coming from @xmath194 , see more detail in ref .
top - quark decays within the standard model are completely dominated by the mode @xmath193 due to @xmath195 to a very high accuracy . in beyond - the - sm theories with an extended higgs sector , top - quark decays can also be done via @xmath194 .
this charged higgs boson has been searched for in high energy experiments , in particular , at lep and the tevatron but it has not been seen so far . to study the higgs boson and new physics beyond the sm at lhc , as a superlative top factory , we need to understand the energy distribution of b - meson produced from top - quark decay .
the dominant decay mode @xmath174 was studied in refs .
@xcite and @xcite . in this paper
we studied the scaled - energy ( @xmath23 ) distribution of b - meson in @xmath175 at nlo in the zm - vfn scheme .
we employed the nonperturbative b - meson ffs determined by a global fit @xcite of experimental data from z factories , relying on their universality and scaling violations @xcite .
comparison of future measurements of @xmath41 at the lhc with our nlo predictions will be important for future tests of the higgs coupling in the minimal supersymmetric sm(mssm ) and it will be the primary source of information on the b - meson ffs .
i would like to thank professor gustav kramer for reading and improving the manuscript and also for important discussions and comments .
i would also like to thank professor bernd a. kniehl for his helpful advices and oppening the doors of perturbative qcd to me .
this work was supported by yazd university and the institute for research in fundamental science ( ipm ) .
25 k. nakamura _ et al .
_ ( particle data group ) , j. phys .
g * 37 * , 075021 ( 2010 ) .
n. cabibbo , phys .
* 10 * , 531 ( 1963 ) ; m. kobayashi and t. maskawa , prog .
phys . * 49 * , 652 ( 1973 ) .
w. bernreuther , j. phys .
g * 35 * , 083001 ( 2008 ) .
j. f. gunion and h. e. haber , nucl .
b * 272 * ( 1986 ) 1 [ erratum - ibid .
b * 402 * ( 1993 ) 567 ] .
atlas collaboration , _ search for charged higgs bosons in the @xmath11jets final state in @xmath2 decays with @xmath196 of @xmath197 collision data recorded at @xmath10 tev with the atlas experiment _ , atlas - conf-2011 - 138 ( 2011 ) a. ali , f. barreiro and j. llorente , arxiv:1103.1827 [ hep - ph ] .
a. czarnecki and k. melnikov , nucl .
phys . * b544 * , 520 ( 1999 ) ; k. g. chetyrkin , r. harlander , t. seidensticker , and m. steinhauser , phys . rev .
d * 60 * , 114015 ( 1999 ) ; i. r. blokland , a. czarnecki , m. lusarczyk , and f. tkachov , phys . rev .
lett . * 93 * , 062001 ( 2004 ) ; phys .
d * 71 * , 054004 ( 2005 ) ; * 79 * , 019901(e ) ( 2009 ) ; r. bonciani and a. ferroglia , jhep * 0811 * , 065 ( 2008 ) .
a. denner and t. sack , nucl .
b * 358 * , 46 ( 1991 ) ; g. eilam , r. r. mendel , r. migneron , and a. soni , phys .
* 66 * , 3105 ( 1991 ) ; c .- p .
yuan and t. c. yuan , phys .
d * 44 * , 3603 ( 1991 ) ; t. kuruma , z. phys .
c * 57 * , 551 ( 1993 ) ; s. m. oliveira , l. brcher , r. santos , and a. barroso , phys .
d * 64 * , 017301 ( 2001 ) .
g. corcella and a. d. mitov , nucl .
b * 623 * , 247 ( 2002 ) .
g. corcella and f. mescia , eur .
j. c * 65 * , 171 ( 2010 ) ; * 68 * , 687(e ) ( 2010 ) ; s. biswas , k. melnikov , and m. schulze , jhep * 1008 * , 048 ( 2010 ) .
d@xmath198 collaboration , b. abbott et al .
d * 58 * , 052001 ( 1998 ) ; + cdf collaboration , t. affolder et al .
d * 63 * , 032003 ( 2001 ) . v. n. gribov and l. n. lipatov , sov
. j. nucl .
phys . * 15 * , 438 ( 1972 ) [ yad .
fiz . * 15 * , 781 ( 1972 ) ] ; g. altarelli and g. parisi , nucl .
* b126 * , 298 ( 1977 ) ; yu .
l. dokshitzer , sov .
jetp * 46 * , 641 ( 1977 ) [ zh .
fiz . * 73 * , 1216 ( 1977 ) ] .
j. c. collins , phys . rev .
d * 66 * ( 1998 ) 094002 .
a. kadeer , j. g. krner , and m. c. mauser , eur .
j. c * 54 * , 175 ( 2008 ) .
s. m. moosavi nejad , `` bottom - hadron production through top quark decay , '' desy - thesis-2009 - 017 ; t. kinosita , j. math .
phys . 3 , 650 ( 1962 ) ; t. d. lee and m. nauenberg , phys . rev .
b * 1549 * ( 1964 ) 133 . a. czarnecki and s. davidson , phys .
d * 47 * , 3063 ( 1993 ) [ arxiv : hep - ph/9208240 ] .
j. liu and y. p. yao , phys .
d * 46 * , 5196 ( 1992 ) [ arxiv : hep - ph/9205245 ] . b. a. kniehl , g. kramer , i. schienbein , and h. spiesberger , phys .
d * 77 * , 014011 ( 2008 ) .
a. heister _ et al . _ ( aleph collaboration ) , phys .
b * 512 * , 30 ( 2001 ) .
g. abbiendi _ et al . _ ( opal collaboration ) , eur .
j. c * 29 * , 463 ( 2003 ) .
k. abe _ et al .
_ ( sld collaboration ) , phys . rev
. lett . * 84 * , 4300 ( 2000 ) ; phys .
d * 65 * , 092006 ( 2002 ) ; * 66 * , 079905(e ) ( 2002 ) .
particle data group , w. m. yao et al , j. phys .
g * 33 * , ( 2006 ) 1 . | we study the energy spectrum of the inclusive bottom - flavored mesons in top - quark decay into a charged - higgs - boson and a massless bottom quark at next - to - leading order ( nlo ) in the two - higgs - doublet model . to extract the result we work in the zero - mass variable - flavor - number scheme(zm - vfns ) using realistic nonperturbative fragmentation functions obtained through a global fit to @xmath0 data from cern lep1 and slac slc on the z - boson resonance .
we study both the contribution of the bottom - quark fragmentation and the gluon fragmentation to produce the bottom - flavored meson ( b - meson ) .
we find that the contribution of the gluon leads to an appreciable reduction in decay rate at low values of the b - meson energy .
it means the nlo corrections are significant . |
* nf : * welcome jerry . let s begin at the beginning , which was not in this part of california .
* jf : * that s correct .
i grew up in a tiny town near the oregon border called yreka : it s `` bakery '' spelled backwards without the `` b. '' yreka bakery is a palindrome and there _ was _ a yreka bakery in yreka .
* nf : * what were your parents doing ?
* jf : * my mother was a housewife and my father , along with his brother , owned a laundry and dry - cleaning establishment there that they and my grandparents founded in the 1930s .
* nf : * were your grandparents born in america ?
* jf : * no , one set was born in the ukraine i think ; i m not sure where the other set was born .
they certainly were nt born in the us , as they all had heavy accents . * nf : * do you have siblings ? * jf : * one brother slightly younger than me .
he s now retired and living in la .
he was an accountant for most of his life .
* nf : * how was school ?
* jf : * school was okay .
i was a dramatic underachiever .
i was nt very interested in school ; i was mainly interested in electronics , so i was into amateur radio , building radio electronics
transmitters , receivers and that kind of thing as a kid .
this was very unusual for someone in yreka .
i was really an outlier , but i thought electronics was fascinating , to be able to talk with people on the other side of the world with no wires .
now , it s just taken for granted . in those days short wave radio was the only way to do it .
when i was really young in grammar school10 to 13i used to build crystal sets all the time . then i graduated to vacuum tubes , transmitters and receivers .
it s very different electronics than today .
vacuum tubes operate at very high voltage . so often while you re poking around trying to see why a circuit is nt working , all of sudden you pick yourself up on the other side of the room because you touched a place at about 400 or 500 volts . today
s electronics run at 5 volts .
i remember bugging the math teacher in middle school to teach me square roots because i needed that to understand some things in this electronics book that i was reading .
* nf : * did you have anybody that you could talk to about this stuff ? * jf : * yes , i had a friend whose father was in amateur radio and knew a lot about electronics , so i could talk with him about it .
my father went to talk to the principal before i graduated high school and asked what he should do with me .
the principal said , `` well he s not going to make it in college .
you might try chico state and when he flunks out you can put him in the army . ''
so that is how i got to chico state .
its claim to fame now is that it s where sierra nevada pale ale beer is brewed .
* nf : * what was your view of this opinion ?
* jf : * i did nt want to go to chico state , i wanted to go to berkeley .
so we struck an agreement that i would go to chico for two years and if i was nt doing too badly , i could consider transferring to berkeley .
my father was right about that .
he was nt often right , but he was right about that . at that time chico state was one of our country s biggest and best known party schools , not big in size , but its reputation as a party school was well deserved .
there were big parties every night .
i used to look forward to summer vacations when i could get relief from all those parties . every night
we drank an enormous amount .
there were no drugs around at that time , but there was lots of alcohol .
when i left for berkeley two years later i was ready to do something more serious , which might not have happened if i had gone directly to berkeley .
* nf : * did you actually go to chico state wanting to learn something specific ? * jf : * i was nt sure what i wanted to be , either a chemist or an engineer .
i think i wanted to be a chemist and i took the elementary chemistry course .
i remember that we were learning how to test for acidity using litmus paper , which is a real ordeal , and i noticed that the engineering students were taking the same lectures as us and they were in the same lab , but their lab was not as intense as ours and they were using some sort of meter .
you put the meter in the solution and it displayed the ph .
i said , `` i like that , '' so i switched to engineering and actually did engineering at chico .
but there was a very , very good physics professor there who got me very interested in physics , so when i transferred to berkeley i decided to study physics .
* nf : * you then spent the next two years as an undergraduate at berkeley .
how well did you do ?
* jf : * i think it actually took me two and a half years .
i was working my way through school .
i had no money .
i did fairly well . those were the days before grade inflation , so i had about a @xmath0 average , which in those days was considered good .
now they get very impatient if you do nt have straight a s , but in those days a s were nt as easy to get .
( see the anecdote _ undergraduate days at berkeley in the early sixties _ in @xcite . )
* nf : * let s move on to your transition from undergraduate to graduate student .
you re at the end of your undergraduate program and youre now deciding what to do . what was your passion of the day ?
* jf : * i wanted to go into physics .
i thought it very interesting and i could nt find anything else i found more interesting .
i never took a statistics course .
* nf : * there was no doubt that you wanted to do it at berkeley ?
* jf : * yes , i loved berkeley , still do .
i like being in the bay area .
however , there was a problem .
in those days there was the military draft . since i had taken an extra semester to go through undergraduate school , i was ineligible for an automatic deferment through graduate school .
and you had to be in school to avoid being drafted into the army .
i thought graduate school infinitely preferable to the army .
so for a while i was worried that i would be drafted because i was classified 1a , healthy and ready to go .
i even went down to the oakland induction center and had my pre - induction physical and so i figured : this is it , i m going into the army .
vietnam was nt big then , so that was not an issue i was worried about .
learning physics seemed to be more fun than the army would be .
one day i received my new draft card they reissued them every year or something like that and instead of saying 1a , it said 2e , which meant student deferment .
so i had a dilemma because i thought maybe it was a typographical error .
the next time i was in yreka , i was torn between either keeping my mouth shut and hoping they would nt discover the mistake , or going up to the draft board and asking them if it was real .
i finally decided i d better find out .
the secretary of the draft board said , `` you are 2e , '' and when i looked at her puzzled , she said , ` well , the draft board decided that since you worked your way through school , it s okay that you took an extra semester to get through.' * nf : * virtue is more than its own reward .
* jf : * i guess so .
also , they re given quotas to fill .
there are many kids in yreka who do nt go to college .
in fact , in those days there were very few , so there were lots of young men not in college whom they could induct .
they did nt necessarily need me to fill their quota .
* nf : * was it hard to get into graduate school ?
* jf : * i do nt know , i think it was , but i was nt very responsible .
berkeley physics was the only graduate department i applied to .
you should apply everywhere , but it was the only one i applied to .
if i had nt been accepted , i would have gone into the army .
* nf : * what was the view of your parents about pursuing graduate studies rather than going back and helping out in the business ?
* jf : * oh , i really knew i was nt going back to yreka .
mack davis , who is a country singer / songwriter , grew up in lubbock , texas .
he was once asked what it was like to grow up in lubbock .
he said , `` well , happiness is lubbock in your rear view mirror , '' and that s the way i usually thought about yreka .
it was a nice place and all , but it was nt the place for me . * nf : * how did your ph.d .
studies go ?
* jf : * they went well .
as things got more difficult my grade point average seemed to go up rather than down and i really enjoyed it ; i loved doing it .
i worked harder and of course there was always the military draft there if you flunked out .
the deferment was good as long as you were in school .
fortunately for me , i did nt flunk out and i really enjoyed learning physics . during the summers i d worked at radio stations , but in the winter when i was at school i worked in the library stacking books , which i did nt really like that much .
my roommate mentioned that there were these great jobs at the lawrence berkeley radiation laboratory .
they did manual pattern recognition on bubble chamber images of elementary particle reactions .
they needed people to scan the film and pick out the particular patterns that they were looking for .
it was a great job , a bit boring , but it paid much better than the library , and so i went up there . that s when i started getting interested in high - energy physics .
the leader of the group was louis alvarez . at the time alvarez
had nt yet received his nobel prize .
he received it later in 1968 when i was a graduate student in his group .
after i got my degree , he and his son were the ones who came up with the meteor / dinosaur extinction theory .
one of the smartest men i ve ever met . * nf : * did you end up working with him ? * jf : * no , i worked with ron ross , one of the professors in his group .
i worked there as a bubble chamber scanner for a while .
then when i had to choose a thesis topic there were two reasons for going into high - energy physics .
one was the alvarez group .
the other one was that in the courses that i took in the first two years my weakest subject was quantum mechanics .
i thought if i went into high energy - particle physics , i would really have to learn quantum mechanics well .
* nf : * were you doing any computing at this stage ?
* jf : * i did nt do any computing
well , actually i did , around 1962 .
the way i started computing is an interesting story .
i was there as a scanner and one of the more advanced physics graduate students would sometimes ask me to do little tasks for him besides the scanning .
one time he asked me to draw a scatter plot .
he gave me a piece of graph paper , a pen and a list of the pairs of numbers .
he said , `` what you do is for each pair of numbers , find the corresponding point on the graph and you put a dot there with the pen . ''
i was doing this for a while and of course i d repeatedly mess up and have to start over again .
one of the other students said , `` you know , down on the first floor they have a thing called a computer and it has a cathode ray tube hooked up to it , and it automatically makes scatter plots .
you can write a program to place the points on the cathode ray tube .
a camera then photographs the tube so you can take a slide of this scatter plot and print it . ''
i thought , _ boy , is that a good idea ! _ i got a book about programming computers and i drew my scatter plots with ease .
* nf : * what were you programming in ? * jf : * machine language and fortran .
fortran was brand new then and the only high - level programming language .
it was very controversial because real programmers did nt program in fortran , they programmed in machine ( assembly ) language .
there was a sign over the entrance to the programming group office that said `` any program that can be written in fortran deserves to be . ''
i guess that s still true today .
* nf : * what was the nature of the hardware ?
* jf : * the first computer that i actually programmed was a vacuum tube computer ( it was nt even a discrete transistor computer ) called an ibm 704 .
it had magnetic core memory .
there was also an ibm 650 with rotating drum memory .
i liked the 650 , even though it was much slower , because for that you could just walk up and use it . with the 704 you had to book time and
wait to get your job run .
the whole thing at berkeley used punch cards .
i did nt see a text editor until i went to slac .
the greatest invention i ever saw was the terminal with the backspace key . with punch cards ,
if you make a mistake , you ve got to throw the card away and start over again from the beginning . in the alvarez group
i was one of those who did most of the programming . in those days , it was considered sissy work to some extent .
real physicists built hardware detectors , particle beams , etc .
programming was sissy work .
high - energy physicists do nt think that way any more because most of them do programming .
but i liked programming much better than building hardware .
* nf : * what were you doing in your ph.d .
studies ?
* jf : * it was part of a large physics experiment in the 72-inch hydrogen bubble chamber , which was the same detector that produced the film i was scanning before .
i studied a particular reaction for my thesis : reactions involving the @xmath1 meson .
* nf : * what sort of hard skills was this calling on , mathematical skills , computational skills ? * jf : * certainly computational skills and understanding the theoretical physics of the time , which did involve some math .
you had to build a program , and that meant figuring out the algorithms to write the program . while i was there as a graduate student i wrote a suite of exploratory data analysis programs that almost everyone in high - energy physics was using .
* nf : * so you were actually writing a statistical package .
* jf : * yes . physicists did nt do much hypothesis testing and things like that ; it was mostly exploratory , automatically making scatter plots , histograms , various other kinds of displays mostly displayed on hardware of the time , which was mostly this line printer output .
_ kiowa _ ( that s the name of an indian tribe ) was a package that i wrote .
it was the standard statistical package in high - energy physics all over the world , for many years .
i also wrote a fast general - purpose monte carlo program called _
sage_. physicists did a lot of monte carlo for simulating particle reactions .
i was still getting enquiries about _ sage _ twenty years later , and i believe that some people are still using it . *
nf : * at some point during your computing activities you came across maximum likelihood .
* jf : * that s probably when i first really started getting interested in statistics .
there was a physicist , frank solmitz , in the alvarez group who knew a lot about statistics .
he d written a little technical report about fundamental statistics for physicists and i thought that was really interesting .
then another guy , jay orear , who was also a physicist , wrote a little note on maximum likelihood model fitting ( @xcite ) .
we were fitting a lot of models and he knew about least squares .
i thought that maximum likelihood was the most elegant idea i had ever seen and it sort of perked my interest in statistics .
of course it was invented by fisher , but i did nt know that ; i thought that jay orear invented it . *
nf : * when did you graduate ?
* jf : * i got my degree in 1968 and then they considered me a good graduate student , so they wanted to hire me as a postdoc physicist at berkeley .
postdocs in those days could run forever and they did for a lot of people .
so i stayed until 1972 in the same alvarez group doing much the same kind of things , different experiments but basically the same stuff . by then slac ( stanford linear accelerator center )
had come online and so i was involved in an experiment that was running at slac while i was at berkeley .
* nf : * had you started interacting with slac ?
* jf : * well , not really , i mean the data was taken to slac , but i never really went down to slac much except to watch the beam .
watching the beam means that you are taking data ; it s a beam of electrons ( at berkeley it was a beam of protons ) and it smashes into matter and then the reaction products come out and they re detected by particle detectors .
there s a huge amount of electronics controlling all that .
so someone has to be in the control room monitoring the electronics to be sure that everything is okay and that you re still taking the data at a reasonable rate .
* nf : * when had slac been set up ? * jf : * slac had been built in the sixties , it may have started in the fifties and it came online in the mid - sixties ( 1966 ) .
this was one of the first experiments at slac .
it was an electron machine , so we were in collaboration with some slac people at berkeley .
our bubble chamber was moved to slac .
the data was taken there and brought to berkeley to be scanned , measured and analyzed .
i did nt spend much time at slac during that period .
* nf : * why were the data going to berkeley ?
* jf : * because that is the way high - energy physics works even today .
there is a lot of data to analyze , it is very labor - intensive , and so you spread the work around and it gets done faster .
* nf : * in other words , distributed computing ? * jf : * in a sense , yes . also , these experiments were very expensive to run , so people like to get together and do it in collaboration . in those days
there were collaborations of tens of physicists , now there are collaborations of dozens of laboratories .
* nf : * why did you move to slac ?
* jf : * well , we had a new director of the research division at berkeley who decided that postdocs should not stay on forever and that three years was the maximum postdoc term .
so he fired all postdocs who had been there for more than three years . that included me , so i had to go out and find a job
. back then , job availability in high - energy physics was cyclic .
there would be a lot of them and then there would nt be many .
this was a time when there were nt many .
i did have a few good opportunities , but they involved moving away from the bay area and i did nt want to do that .
so frank solmitz , the physics statistics guy , came up to me one day in the hallway and said , `` there s a position at slac leading a computer science research group and they were asking me who might be a good computing physicist for that and i mentioned your name .
are you interested in exploring it ? ''
i thought it was nt really for me but i could explore it .
so i went down and i interviewed .
first i interviewed with all the directors and all the group leaders at slac , then i interviewed with all of the professors in the computer science department on campus .
originally they wanted to get a famous computer scientist to run that group , but they could nt find one that they liked and who liked them , so they decided to get a computing physicist , which is why they landed on me .
after i returned from interviewing i figured that was it .
it was a fun experience , but i did nt think i wanted it and they did nt want me .
then i got a call a week or so later saying , `` there s been more than a little interest in you .
what do you want to do ? ''
i said , `` i think i d better talk to the people in the group before i do anything else . ''
i went and talked to the people in the group .
they were really good people , so i thought , _ why not _ ?
so i went down to slac to lead this computation research group .
it was set up by bill miller , who initially established the computing facility at slac .
they wanted him to build up the computing center so he would only come under certain conditions .
one condition was that he be made a professor in the computer science department .
another condition was that he would be able to have his own computer science research group at slac .
slac had a lot of physics research groups but he would have his in computer science and that was this group .
he eventually became provost of the university ( stanford university ) , so that position was open and that s where i went .
* nf : * how were things set up ?
* jf : * he had a lot of bright people there .
a number were in computer graphics , which was in its infancy in those days .
he had set up a really state - of - the - art computer graphics facility , including movie - making equipment worth millions of dollars , which was a lot of money in those days .
it was really state of the art .
there were people doing research in other areas of computer science , and a few pure service types doing job - shop programming for the physicists at slac ; overall , about ten people in the group .
* nf : * so you had the sort of technology advantage that the bell labs statistics group had rather later on with their workstations .
* jf : * yes , this was a fantastic facility . also , slac was a physics lab and high - energy physics labs had more computing than anybody else except for weapons laboratories .
i had access to the computing facilities at slac , including their mainframe computing system . very few statisticians had access to that kind of computing at that time or even fifteen years later .
* nf : * what did the job involve ?
* jf : * the job involved mainly running the group as an administrator and then doing my own research .
i think they expected me to do half and half : i did about one quarter administration , three quarters research .
i arrived there in early 1972 , commuting from berkeley for the first six months .
also , i was asked to teach an elementary computer literacy course in the computer science department .
it was a course on algorithms , data structures and computer architecture .
i knew some of those things a little bit , but in order to teach the course i had to learn them all in detail .
it was one of the most valuable courses i ve ever taught in terms of what i learnt .
i still use most of it in my work today .
the research that i wanted to do was in pattern recognition . even when i was a student and then a postdoc at berkeley , i was interested in data .
id written some analysis packages , i d done monte carlo , and i d written a program to do maximum likelihood .
my interest in data worked out well because most other physicists were more interested in building new equipment at that time , whereas i was interested in analyzing the data and that is what got me into computers .
i loved computers .
* nf : * what did you try to do with pattern recognition ?
* jf : * it was called pattern recognition then ; it s called machine learning now . sort of basic pattern recognition , like nearest - neighbor techniques .
i d read the cover and hart ( @xcite ) paper and i was interested in clustering and in general statistical learning , but it was nt called that then .
the closest name then was `` pattern recognition . ''
* nf : * finding groups in data ? * jf : * yes , finding groups in data , using data to make predictions , that kind of thing .
i did nt have a clear - cut research agenda at that particular time .
i was just leaving berkeley where i d mainly done physics except for the other sort of statistical things , so i had nt really developed a research agenda .
i m not sure i ever had one .
* nf : * i understand that the group that you were involved with there had some extraordinary people .
* jf : * yes , it did . when i came , it was common then , and may still be , that in the ( stanford ) computer science department professors were paid half their salary from the department and expected to go out and raise the other half externally .
one way they could do that would be to work in other places . in our group we often had computer science professors working part time . when i came , gene golub was halftime in the group .
and we had two visionaries , harry sahl and forrest baskett .
harry was there when i came .
forrest joined later .
this led to some remarkable developments .
( see the anecdote _ building the first graphics workstation _ in @xcite . ) * nf : * just after you moved to slac you started collaborating with john tukey .
* jf : * yes , my predecessor , bill miller , was close friends with john tukey , so he d invited tukey to come out during his sabbatical because , as we all know , john was very interested in graphics and he was especially interested in motion graphics .
our facility was one of the very few places you could do motion graphics .
when i arrived at slac everyone was excited that this guy was coming , not because he was a great statistician , but he because he was well known in computer science for having invented the fast fourier transform .
they were really excited , and i d never heard of him . * nf : * so when john came up you did not actually have a research project in mind ? * jf : * no .
i talked to him and he told me what he was doing , what he was interested in , and i found it very interesting .
we just hit it off .
he worked on the graphics , i worked a little bit on the graphics but not a lot .
i would watch what they were doing with the graphics rotating point clouds and isolating subsets , saying , `` okay , let s just look at these , '' and so on trying to visually find patterns in data .
john was mainly working with a programmer in our group .
* nf : * john never programmed , himself ? * jf : * not to my knowledge , at least not code that ever ran on a computer .
he wrote out his thoughts in a kind of pseudo - fortran , but he never actually sat in front of a terminal to execute code , as far as i knew .
( see the sample of tukey s research notes in @xcite . )
* nf : * what sort of ideas was he having at that time , point cloud rotation and so on ? * jf : * well , if you see the prim-9 movie , that s the product and those were the ideas he had .
it was basically integrating the idea of rotating point clouds in arbitrary orientations .
he was very interested in human interfaces and he developed some really slick controls , especially given the crudeness of the equipment he had to work with .
i was watching what he was doing and he would iterate to an interesting picture and so i started to think : _ what makes the picture interesting ? _ and i would discuss this with him .
he said , `` it seems that the pictures we like the most are the ones that have content ; they have a lot of small inter - point distances but then they expand over the whole thing . ''
when i was at berkeley i had been working on optimization algorithms and i thought , _
what if we defined some index of clumping and then tried to maximize it with an optimization algorithm ? _ that was basically the beginning of projection pursuit and we interacted on that .
so i was off doing the analytical algorithm and john was doing the graphics . * nf : * what was john s interest here ?
he was nt actually trying to tackle a scientific problem to do with physics ?
* jf : * well he thought it would have a big application in physics because physics has inherently high - dimensional data with a great deal of structure .
it was nt like the sort of diffuse data that comes from the social sciences : data from physics have a very sharp structure .
in fact , i think the data set that s illustrated in the movie is a high - energy physics data set .
so his vision was that it could be used for high - energy physics , but i think he was certainly thinking about the bigger picture . [
cols="^,^ " , ] i think he was there four months .
when he came back later for a little while , i said , `` john i think we ought to make a movie of this , '' since we had a lot of movie - making equipment .
my predecessor bill miller was a genius at raising money .
he had a graduate student who was interested in graphics .
the student was very smart and wanted the best of everything , so he got the best of everything .
he knew how to handle the movie equipment , so he made the film just pointing a camera at the screen with john there talking .
so then we had a film and then no one wanted to edit it .
a new member , sam steppel , had just joined the group and i asked , `` would you like to do the editing ? '' and he said , `` oh yeah . ''
it turned out to be a big job .
anyway , that was the result of john tukey s first trip to stay with us at slac .
we stayed in contact throughout the 1970s and he came back again for his next sabbatical seven years later . *
nf : * how did you find interacting with him on the original projection pursuit paper ? * jf : * he was very full of ideas and he was very stimulating .
we seemed to talk the same language , to think about things the same way .
his approach was operational : here s the task , here s the problem , how do we approach it , how do we get it done .
he did nt seem to be interested in fundamental principles ; he probably was , but he never said so . * nf : * a very engineering approach . * jf : * very engineering , that was always his approach .
he always delighted in slightly puzzling you by hiding , not telling you the fundamental reason for whatever he was doing , what lay behind it , what were his reasons .
he would come to you and say , `` okay , here s a procedure : you do this , then you do this , then you do this , then you do that . ''
i was young and brash at the time so i would say , `` john , okay , i understand that , but why would you do this and this ?
why is that a good idea ? '' he would repeat , `` well , you do this , then you do this , then you do this , then you do this , '' and i d say , `` john , but why ? '' it would go back and forth like that , him acting like there was no guiding principle .
i guess i was persistent enough that he would finally get exasperated and say , `` oh well , '' and lucidly enunciate the guiding principle ; he had it all the time , he just did nt want to reveal it , at least not right away .
his main thought was he would evaluate a procedure by its performance , not by its motivations .
he was nt interested in : _ is this a bayesian procedure with a particular prior ? is this a procedure that s optimal in some sense ?
_ he did nt come from that perspective .
he would say , `` all right , you ve got a procedure , tell me the operations on the data , the explicit operations .
i do nt care where it comes from , i do nt care what your motivation is ; you tell me the operations that apply to the data and i ll tell you whether i think it s a good idea or not . ''
that s the way he thought about things .
* nf : * do you think he was mentally checking this against a hidden set of principles or seeing how it sat with his instincts ?
* jf : * i do nt know whether he always had a guiding principle or he d make one up so i d stop asking .
i wrote up the first draft of the projection pursuit paper , he edited it and then we discussed it .
my first journal publication of any sort in statistics was the projection pursuit paper with john ( friedman and tukey , @xcite ) .
this is the only paper i have ever submitted and had accepted immediately without revision , and i thought , _ this is really neat , i like this field_. but it s never happened since .
* nf : * you did some follow - up work with him at slac .
* jf : * yes , he came back his next sabbatical in the early 1980s .
i think at that time he was on his way to hawaii because a cousin or somebody was getting married , and elizabeth finally convinced him to take a vacation there on the beach .
so he stopped by stanford and we worked together .
he was very impressed with the fact that the home he was staying in on campus had a swimming pool ; it was the house of a professor who was on sabbatical .
so he would nt come in the morning ; he would spend his mornings sitting by the pool , maybe swimming as well , writing out ideas lots of ideas about how to analyze high - dimensional data , usually writing in cryptic words or pseudo - fortran .
then he would bring them in later in the afternoon and ask our secretary to type them up .
this happened every day .
later , werner ( stuetzle ) and i would take a look at them and sometimes discuss them with him .
he finally took off for his vacation in hawaii and the notes stopped .
a few days later , packages of notes started arriving in the mail , every day another package from hawaii .
he was thinking on the beach instead of at the swimming pool .
i ve still have many of these notes .
after john died , there was an issue of the _ annals _ that had a long article about him by david brillinger ( @xcite ) , and werner stuetzle and i wrote a shorter article ( @xcite ) talking about his graphics work and our experiences with him in his graphics work . at that time we thought maybe we should get the notes together ,
take a look at them .
there are probably a tremendous number of ideas there that are still revolutionary by today s standards in terms of data analysis , but this is one of those things you do when you have time .
* nf : * going back to your own personal research , it seems that it was becoming more statistical .
* jf : * it was .
i was interested in pattern recognition in the general sense and among the more popular methods of the time were nearest - neighbor methods and kernel methods .
cover and hart had shown that , asymptotically , the nearest - neighbor classification method reaches half the bayes risk just with the nearest neighbor .
of course , at that time we did nt appreciate the difficulty of becoming asymptotic in high - dimensional settings . at the time people were very excited about it and i thought , _
well , if we are going to use this approach in applications with bigger data sets like those in high - energy physics , we ll need a fast algorithm to find nearest neighbors in data sets_. at the time slac experiments generated tens of thousands of observations , not millions like now , but tens of thousands .
the straightforward way to compute near neighbors is typically an @xmath2-squared operation : for each point you have to make a pass over all the other points .
so i started working on fast algorithms for finding near neighbors , without too much success .
then i met jon bentley , a student of don knuth s .
he had some really clever ideas based on what he called _
k - d _ trees , and so he and i started working together with another student , raphael finkel , on trying to develop fast algorithms for finding near neighbors .
so probably one of the papers that i am best known for outside statistics is that paper : fast algorithms for finding near neighbors ( @xcite ) .
then jon went off to graduate school at the university of north carolina .
after that he went on to do great things and became very famous in computer science . the whole _
k - d _ tree idea is considered a very important development in computational geometry , and john invented it , an unbelievably bright man .
another interesting aspect is that that s what got me into decision trees , because the _ k - d _ tree algorithm for finding nearest neighbors involved recursively partitioning the data space into boxes .
if you wanted to find the nearest neighbors to a point , you d traverse the tree down to the box containing the point , find its nearest neighbor in the box and backtrack up and find its nearest neighbors in other neighboring boxes using the tree structure .
that was the algorithm .
i was thinking : okay , if you want to find nearest neighbors , that s fine , but suppose the purpose of finding the nearest neighbors is to do classification , maybe there would be modifications to the tree - building that would be more appropriate for nearest neighbors in that context .
so it occurred to me that in the nearest - neighbor algorithm you could recursively find the variable with the largest spread and split it at the median to make boxes .
why do nt we find the variable that has the most discriminative power and split it at the best discriminating point ?
so i came up with that paradigm to find the nearest neighbors .
then it occurred to me that you did nt need the nearest neighbors at all ; you could use the boxes ( terminal nodes ) themselves to perform the classification .
* nf : * when was this happening ?
* jf : * probably around 1974 , before i went to cern .
that was my initial thinking about what eventually became cart : it came from the recursive partitioning nearest neighbor algorithm to get the tree structure .
somewhat later i joined with leo breiman , richard olshen and chuck stone who had been independently pursuing very similar ideas .
oh , i forgot to mention that when i first joined the computation research group in the early 1970s , gene golub came to me one day and said , `` i m going on sabbatical next year , which means that i wo nt be here and i m worried that if you have an empty position for a year it might not be there when i get back .
so i think you should fill it with someone and i know just the ideal guy .
his name s richard olshen and he s in the statistics department . ''
so i hired richard half time .
that was in the early days that i was working on trees .
i was talking to richard and he asked , `` what are you doing ? '' `` well , i m working on this recursive partitioning idea .
'' richard got very interested in it and he has made great contributions to tree - based methodology over the succeeding years . * nf : * after a few years at slac , you decided to take a sabbatical at cern .
did you have a family at this stage ?
* jf : * yes , i had a wife and a three - year - old daughter at that time , and we all went to cern , in geneva .
it was natural that when physicists took a year off they went to cern .
it was nt an official sabbatical , i just decided i wanted a year away and so i asked for a leave of absence .
i was a staff member , but i was nt a faculty member .
intellectually , it was not super stimulating .
i was in the computer group which was called data handling and it was a big group at cern that had the computers .
the professional thing i did was to work on adaptive monte carlo algorithms .
what i mainly did was eat their food , drink red wine and dine at a lot of michelin three - star restaurants , which is what i mainly remember .
cern was a lot of fun .
slac was quite an intense place , whereas cern was much more laid back at that time .
* nf : * did you visit any other groups while you were at cern ?
* jf : * yes , i did , which turned out to be very important for me .
when i was at cern i got a letter from john tukey saying , `` there s this fellow i know in zurich at eth , peter huber ; he is interested in these projection pursuit kinds of stuff .
you should go and visit him . ''
so i went to zurich and found my way from the train station to eth .
i d never met peter or anyone else from eth , so i was standing there in a hallway , and a guy came up to me and asked , `` can i help you ? ''
i guess he knew i spoke english , maybe it was written all over me .
i said , `` yes , i m trying to find peter huber .
'' he turned out to be andreas buja , who was peter s student at the time . on that trip
i also met another of peter s students , werner stuetzle .
we had a strong collaboration throughout the early 1980s when he came to slac and stanford .
i think andreas also visited slac a couple of times .
both are unbelievably smart guys .
* nf : * returning to your time at slac , you d started attending interface conferences and meeting people * jf : * yes . i met leo breiman and chuck stone at an interface meeting in 1975 .
leo gave a talk about nearest neighbor classification or something and i was working on these fast algorithms at the time , so i raised my hand at the back of the room and said , `` we ve been working on some new fast algorithms for finding nearest neighbors . '' after the talk leo looked me up .
he was very interested and we started talking , but that was pretty much it .
but then he sent me a letter while i was at cern saying that he was organizing a meeting in dallas in 1977 ; he called it a conference on _ the analysis of large and complex data sets_. leo was another visionary ; he saw into the future of data mining .
he invited me to give a talk there .
i d never been to dallas and so soon after i got back i went to that meeting , and that meeting to a large extent changed my life professionally .
i met larry rafsky there , with whom i later collaborated , and i also met bill cleveland . * nf : * how did this conference change your life ?
* jf : * because i met leo again .
* nf : * we ll talk about leo shortly .
you did some work with larry rafsky around this time .
* jf : * yes , we started talking about some of our mutual interests in computational geometry ( near neighbors ) .
this led to the work in the late 1970s , early 1980s on using minimal spanning trees for multivariate goodness of fit and two - sample testing , leading also to general measures of multivariate association .
two _ annals _ papers came out of that ( friedman and rafsky , @xcite , @xcite ) .
i was also refining the recursive partitioning idea , extending it in various ways , and i worked with larry a bit on this as well .
he was a very bright guy with lots of ideas .
i learned a lot from him .
* nf : * let s bring the background murmurs about recursive partitioning to the foreground and talk about cart .
how did this celebrated collaboration come about ?
* jf : * after larry and i wrote the two papers using minimal spanning trees , we started working on the cart idea .
richard olshen was at uc san diego at this time ( mid-1970s ) and he made trips every once in a while back to stanford , and he would come out and visit me at slac .
sometimes i would tell him about the more recent work on trees .
he d done some nice theoretical work with lou gordon ( louis i. gordon ) , a former stanford professor who was working in industry at that time .
i told him how we were extending decision trees and he said , `` it sounds a lot like what leo breiman and chuck stone are doing down in la . ''
he tried to explain to me what they were doing and i did nt quite get it ; and apparently he was trying to explain to them what we doing and they did nt quite understand either . finally , chuck called me and we had a long discussion .
we d been working totally independently , but there was a huge amount of commonality in what we were doing .
so i guess it was leo who finally suggested that we have a meeting down in southern california . they were both consultants for a company called technology service corporation that was operating on government contracts , mostly environmental things
leo was basically a full - time consultant there and chuck was also a consultant .
in fact , some of the technical reports that they wrote then are the classic articles on trees .
so , larry and i and chuck and leo , we went down there ( richard was nt there ) and had a meeting at tsc .
we talked about how very exciting it was and that there was a lot of commonality in our respective approaches .
there were some differences , and we discussed which ones seemed best .
then leo said , `` hey , i think we ought to write a monograph . ''
we would never get something like this published in a statistics journal ( of the day ) .
so we set off to write it , and that s how the monograph was born ( @xcite ) .
* nf : * as i recall , there was other work on recursive partitioning going on about this time .
* jf : * well , it s one of those ideas that s continually re - invented .
everybody who re - invents it thinks this is their `` nobel prize '' moment .
there was the work of @xcite , in the early 1960s at the university of michigan social science center ; they did trees .
then there was ross @xcite who was doing what he called the iterative dichotomiser 3 ( id3 ) algorithm , a crude tree program , at about the same time .
later he did c4.5 , which turned out to be very similar to cart , although there are a few differences .
we take pride in the fact that cart came ten years earlier than c4.5 , but it was quinlan and the machine learners who popularized trees .
we did cart and it just sat there : statisticians said , `` what s this for ?
what do you do with it ? ''
* nf : * and you d also implemented the software and made it available .
* jf : * yes , we d made it available .
then we got the idea of trying to sell it and that s how our little company got started .
* nf : * first , let s talk about your long collaboration with leo . this was the beginning .
* jf : * right , it started with cart because we were trying to write the software .
i had written the initial software , but leo had a lot of good ideas about what should be in it and how it should be structured , the user interface etc . , and
so we were collaborating on that . in the meantime , leo left ucla and became a full - time consultant .
* nf : * he was a probabilist at one stage .
* jf : * he was a probabilist , he used to say _ probobobilist_. then he came back to academia in 1980 and joined the statistics department at berkeley and , at the same time , chuck came up to berkeley . * nf : * would you say leo was an unusual appointment at berkeley for that time ?
* jf : * yes . he had solid mathematical credentials .
he was like tukey in this sense : he could do this super empirical stuff but he was also very strong in math , so they could nt say that he was doing methodology because he could nt do math .
i have no idea why they hired him , but my guess would be they wanted to start getting into the computer age and they brought him in .
he bought their first computer , a vax , installed it , and did its care and feeding for a long time , so it was an incredibly wise appointment from that perspective , as well as many others .
* nf : * how did the collaboration go ? * jf : * we d started the collaboration with cart , we d decided to write the book , and we d parceled it up into different parts
. then leo says , `` if we write this program called cart and decide to sell it and we sell a thousand copies at a hundred dollars each , you know how much money that is ? ''
so we decided , okay , we would form a company , california statistical software , and try to sell cart .
so we had to have a product .
leo was at berkeley at that time , so we started a pattern that persisted for roughly the next ten years .
every thursday i would go up to berkeley .
i would leave here around 10 am , get up there around 11 and park on hearst avenue .
leo would block out the whole day ; nobody else would come to see him for that day .
we would go to his office and start working . around noon
he d say , `` jerry let s go have lunch , ''
so we d go over to the same place every time , a crpe place over on hearst avenue .
we d have usually the same spinach crpe with sour cream , and an espresso .
then we d go back to his office and work , punctuated with me running out to feed the parking meter on hearst .
it was all conversational ; we were nt sitting there writing or typing into a computer , we were just discussing the whole time . typically , around 5.30 or when the progress seemed to be slowing , leo would say , `` jerry let s go have a beer , '' so we d go down to _ spats _ , which is a pub on shattuck avenue .
after we d had a few beers leo would say , `` jerry let s go to dinner , '' so we d go to one of berkeley s better restaurants and have a nice meal .
then leo would go home and i d drive back down to palo alto . that was the routine every thursday for a very long time .
* nf : * what was his approach to problems ?
* jf : * he was like tukey : `` do nt tell me the motivation , tell me what you do to the data . ''
he was totally algorithmic .
there was no obvious sort of fundamental principle like : _ this is a bayesian procedure with a particular prior_. it was never that kind of thinking , starting from any kind of guiding principle ; it was just what it made sense to do with the data .
* nf : * would you categorize this as the computer science way of tackling data rather than the statistical way ?
* jf : * i would have then .
* nf : * in the sense that what you are doing is looking at a specific data set and you do nt know whether what you ve done is going to work on any other data set ? * jf : * well , we generally were nt working on specific data sets , we were trying to develop methodology for classes of problems .
it was like developing cart : cart could be used on a wide variety of data sets , so could ace ( alternating conditional expectation ) ( @xcite ) , so could curds and whey ( breiman and friedman , @xcite ) .
we were thinking methodologically .
in other words : _ problem .
i ve got data , there s an outcome , there are predictor variables , the data is of a certain kind . now how do we make a procedure that can handle this problem ? _ i do nt think we ever actually analyzed a specific data set together , except for examples that we used in papers to illustrate the methodology . analyzing a data set where the interest was not in how well the method did but in the answer that you got from the data set , we both did a lot of that as well .
* nf : * what motivated ace ?
* jf : * the idea was to simultaneously find optimal transforms .
there were all these heuristics and rules for transforming data in the linear regression problem : do you take logs , or do you take other kinds of transformations ?
in fact , i think box
cox was a sort of automated method for trying to find transformations from a parametric family of functions .
we were involved with smoothers , so we thought about how we could automatically find good transformations without having to restrict them to be from a parametric class of functions , just see if you could estimate an optimal set of transformations .
* nf : * `` optimal '' in what sense ?
* jf : * optimal in the squared error sense of course under a smoothness constraint , otherwise there were an infinite number of transformations that would fit the data perfectly .
so you had to put in a smoothness constraint , which we did explicitly by using smoothers in the heart of the algorithm .
i remember one of the thursdays when i went up to berkeley , leo asked me , `` if i have two variables , how do i find the function of one of them that s maximally correlated with the other one ? ''
i said , `` well , if you do a smooth , you take the conditional expectation of one of them given the other one , okay ? '' that does nt necessarily maximize the correlation , so we started thinking : _ okay , what if we did it one way and then , given that curve , smooth that against the other one ?
_ later we went back to leo s house where he had an apple 2 .
he programmed it in basic , just the simple bivariate algorithm .
he simulated data from a model where the optimal transformation in both cases was the square root .
the apple 2 was not a very fast machine , so we could watch it iterate in real time , displaying the current transformations at each step .
starting from linear straight lines , we saw the transformations begin to become more and more curved with each iteration until they converged .
it was an exciting moment for us .
so we developed that idea and then leo got very excited about the theory .
he never took theory very seriously but he loved to do it , so he looked at the asymptotic consistencies and things like that , and we had a great time . in the early 1990s i went on sabbatical for a year and we did nt collaborate then , but it picked up again in the mid-1990s .
leo called me one day and just said , `` jerry , i d like to work with you again , '' and we did nt even have a specific project to work on .
i went up to berkeley and we kicked around what we could work on .
i said , `` well , one problem i ve been churning in my head but havent gotten very far on is multivariate regression , where you have multiple responses . ''
so we started kicking that around and that led to the curds and whey paper , which was a discussion paper at the royal statistical society .
this collaboration was nt quite in the same mode as before .
i would nt go up to berkeley nearly as much because infrastructure had developed so that it was possible to work apart productively and so we basically did it through e - mail .
the idea was motivated by my familiarity with pls ( partial least squares ) .
pls had a mode where it had multiple outcome variables as well as multiple predictor variables .
the one - outcome - variable case was just a special case . in the work that i had done with ildiko [ frank ] to try and understand pls ( see below ) , we only treated the single outcome case .
i wanted to try to understand the multiple outcome procedure to see if one could find a more statistically justifiable approach .
so leo and i worked on that together and that was great fun . in this paper
we reversed roles .
generally , in our collaborations i concentrated on the methodological part and the computing .
leo would usually do the theory . in this paper our roles
were reversed : leo wrote the program , leo had the data , and i worked out the theory .
* nf : * why curds and whey ?
* jf : * i ll retell the story i told in the memorial article i wrote about leo .
i came up with the name ace .
i liked it a lot but leo hated it , absolutely hated it .
this was one of the afternoons after we finished and we d gone down to _ splats _ for a beer and we were still discussing this .
leo did nt like it and i liked it , so we were going back and forth .
and then out of nowhere leo said , `` okay jerry you ve got it , it s ace . ''
it was most unusual for leo to yield so easily .
he usually stuck to his guns and so did i. i looked at him in a puzzled way , like , _ that was too easy _ , and he said , `` look across the street , '' so i looked across the street and there was a hardware store with this big red sign , _ ace_. when we gave the invited _ jasa _ paper in 1987 , leo brought a bunch of bags from ace hardware that had a big ace on them and distributed them around to the audience . later on when we did the multiple response multivariate regression work , we had another argument about how to name that procedure .
leo proposed curds and whey , which i really did nt like , but i felt that since he had conceded on ace i would concede on that .
it was leo s thinking about the fact that we were separating a signal from the noise , the good stuff from the bad stuff , separating the curds from the whey or the other way around , i guess , in cheese manufacturing .
that collaboration was a couple of years , maybe three years .
i think to some extent our interests separated at that time .
they tended to be concerned with very similar problems .
he did the nonnegative garrotte and then got into bagging and i was getting into boosting at that time working with rob [ tibshirani ] and trevor [ hastie ] .
both approaches were based on ensembles of trees , but from different perspectives .
i knew what he was doing , but we did nt have constant interaction and involvement .
when we got together we always had a good time .
* nf : * talking about leo has had us leaping through the decades .
let s return to the period when you were still full time at slac .
had you met anybody from the statistics department at this stage ?
* jf : * no , not at this stage .
i did nt start interacting with the statistics department until the late 1970s .
* jf : * i was hanging out around the department for seminars , but i had no official position .
so brad [ efron ] asked me to teach a course .
* nf : * did you think you were doing statistics ? * jf : * well yes , i knew the stuff with rafsky was statistics , it was hypothesis testing . that s what i taught in the course .
it s probably as close as i ve come to classical statistics .
the minimal spanning tree was not classical statistics but the rest of it was . that brought me closer to the department .
while i was at slac i was nt on the faculty there .
i was just a staff member , which meant i could nt write proposals and submit them to nsf or other agencies , department of energy , or others who might sponsor my kind of work .
slac was sponsoring it and that was wonderful , but sometimes i really could have used a little more money to do things .
so i wanted to write proposals and for that i needed to be some kind of professor .
paul switzer was chair of the department at that time , so i went to him and said , `` is there any way you could make me something like a consulting professor of the department , some official thing ? this will allow me to write grants and reports on behalf of stanford university . ''
he said , `` okay , we ll try it . ''
so all the paper work was gotten together and submitted to the administration , letters and everything .
it came back and paul said , `` sorry we ca nt do it .
we re not making any more consulting professors ; there is some political thing going on that has nothing to do with your case , but they are not doing consulting professors .
however , they did say that your folder looked pretty strong , so why not try for a regular professor ? '' and so paul did and it worked .
paul probably did the lion s share of the work on it because he was chair . that s how i became a professor
. * nf : * as well as having a job at slac ?
* jf : * i became a half - time professor and half - time at slac instead of full time .
* nf : * so this was effectively your formal entry into the statistics community .
did you find yourself welcomed ?
here s mainstream statistics flowing along and this guy surfs in on a wave from a merging stream with no statistics background whatsoever , but with lots of skills and different ideas about how to approach data . was this a great issue for you ?
* jf : * yes in general , but certainly not at stanford because they hired me .
i always felt very welcome in the department .
but i do nt think the more general statistics community understood what motivated me .
i recall once colin mallows listening to one of my talks , and he said afterwards , `` boy this is really fascinating , but it s not statistics , '' and i think that was the general feeling , that what i was doing was perhaps interesting but not statistics .
where s the math ? where are the usual trappings of research in statistics ?
it really was nt that sort of stuff , with the possible exception of the minimal spanning tree work .
so in that sense , i do nt think there ever was any hostility of any kind , just that people were puzzled : how was what i was doing related to statistics ?
* nf : * and yet what you were really doing was what you described earlier : you and john tukey thinking the same way , you d have an idea about how to attack something and you d see how it worked on the data .
your work was nt being informed by fundamental principles or was it ?
* jf : * i think that had more of an influence on me many years later , and john thought i d sold out .
he really thought i was trying to think about fundamental principles , whereas i was developing things and using elegance of the algorithm as a criterion .
john had a real distaste for that .
* nf : * do you feel that you had developed some sort of a canonical way of tackling the sorts of problems that you approached ?
* jf : * probably , but i ca nt think of it right now .
i operate in the model of a problem solver : here s a problem , i have a certain set of tools and skills that i use , and so that directs everything .
probably there is a great deal of commonality simply because my skill set is limited , but i do nt think i consciously think that way . *
nf : * suppose a young person came to work with you and you treated that person the same way as john tukey used to treat you : you do this , you do this . if you got pushed would you make up a principle or would you actually be able to find a principle ? you suggested earlier that maybe john made the principle up to shut you up . *
jf : * a heuristic principle perhaps , i do nt think i could come up with a deep theoretical principle , or maybe i could if i thought about it .
* nf : * joining the department put you into contact with mainstream statistics and statisticians and you started going to more stats conferences ? how was being in that department changing what was happening ? * jf : * well , i started becoming more conscious of statistical principles .
i do nt think it changed the way i approached problems a lot .
i recall a statement of john rice s when he was asked whether he was a bayesian or a frequentist and he said , `` i m an opportunist . '' and that s how i view it : _ here s a problem .
how do we solve it ? _
i will try to attack the problem from any direction i m capable of .
* nf : * you were also coming into contact with a remarkable group of statisticians in the department , who were doing extraordinary things .
* jf : * i think subconsciously that really shaped my thinking a lot . that s maybe why tukey thought in later years i was selling out .
i did think about principles ; i think they were in the back of my mind , informal principles that i did nt apply formally .
* nf : * did john ever visit you once you had moved into that department ?
* jf : * yes ,
oh yes , at least a few times .
i do remember one time we were driving along campus drive and i said , `` you know , john , now i m in a statistics department and officially in statistics , maybe i should really go and learn basic statistics , theoretical statistics , all the usual stuff .
'' john looked at me and went : ( _ raspberry sound _ ) .
whenever you said anything to john , presented an idea or whatever , john did nt tend to lavish praise , that was nt his style .
so if he sat still and listened to you quietly , you knew he really liked it .
if he had doubts about it , he would nt say anything , but you d see his head going slowly back and forth ; and if he really did nt like it , he d interrupt you by giving a thumbs down and blowing a raspberry . so that s what i got when i asked him whether i should learn statistics
i m not sure he was exactly right and over the course of the years i did learn some traditional statistics with the help of my friends , colleagues and students , which i think helped me a lot . * nf : * you developed more strong collaborative work at stanford .
what was the first one ?
* jf : * around 1981 , the department had an opening for an assistant professor and i think werner [ stuetzle ] had just got his degree .
i said , `` i know this really smart guy that i met at eth .
i think i can pull it off so we pay him half time with my group , do you want to hire him ? ''
they thought about it and , to cut a long story short , they said , `` sure . ''
i convinced my bosses at slac that we could do it , so we hired werner half time at stanford and i had werner half time in my group at slac .
i started my collaboration with werner , which was very profitable intellectually and great fun over the years .
* nf : * what sort of things were you doing ?
* jf : * well , we started a graphics project .
werner had worked with peter huber on graphical techniques , as peter was very interested in that .
so we got some money from the office of naval research and started to put together a graphics workstation .
we felt : it s been ten years since prim 9 , the technology has advanced dramatically , let s see what we can do now .
so we jointly worked on that ; we called it the orion project and it was great fun .
( see the anecdote _ the orion project building a second graphics workstation _ in @xcite . )
* nf : * your full - time work with slac had been a very exciting period of your life .
now you had moved across to the department of statistics , how long did the interactions with slac continue ? * jf : * they tapered off a little bit because i was only half time there and running the group was about a quarter - time exercise , so i had less time to work on slac types of things .
but it was still very valuable to be in that group .
i still had access to a lot of resources that i would nt have had otherwise .
* nf : * how did this change your sources of inspiration for things to work on ? * jf : * i was always interested in what leo called large and complex data sets ( now called `` data mining '' ) : data that was collected not necessarily for the purpose for which you are using it ; it has mixtures of all kinds of variables ; the experiment was nt designed ; it was usually observational data .
i guess it s a kind of data that i first encountered in physics , moderately high - dimensional , a fair amount of data , the number of observations usually considerably larger than the number of measured variables .
i was always interested in developing general - purpose algorithms where one could pour the data in and hope to get something sensible out without a lot of labor - intensive work on the part of the data analyst .
* nf : * jerry s search for pattern ?
* jf : * yes , i guess a generalized pattern search of data , usually focused on prediction problems .
* nf : * looking forward from your arrival at slac , first there was projection pursuit where you were looking for groups in high - dimensional data ?
* jf : * i think i was associated with four projection pursuit papers . one was the original tukey paper ( @xcite ) , then there was a regression paper with werner stuetzle ( @xcite ) , then i wrote another follow - up paper ( @xcite ) in the original tukey style , and one with werner on density estimation ( @xcite ) . in the mid-1970s
i began work on trees which carried through to cart .
then i went back to trees later in the 1990s when the various ensemble methods were coming out .
ensembles of trees seemed especially appropriate for these kinds of learning machines because trees have a lot of very desirable properties for data mining .
trees just have one problem : they are not always very accurate .
so the ensembles of trees cured the accuracy problem while maintaining all of the previous advantages ; they are very robust , they can deal with all kinds of data , missing data , and that s the kind of thing i was interested in : off - the - shelf learning algorithms .
you could never do as well as a careful statistician or a careful scientist analyzing the data very painstakingly , but it could give you good first answers , that was the idea . that s basically what drove me .
research interests are a random walk .
you get an idea and you pursue for a while
. it may be similar to what you were working on before or it may be in an entirely new direction .
you work on it for a while until you get stuck or find something more interesting .
i tend to have these problems that i would like to solve and ca nt solve immediately .
i put them in the back of my mind and then when i m reading or hearing talks or every once in a while someone says something that may have nothing to do with what s in the back of my mind , it will trigger something : _ ah ha !
there s an idea that i can try for this problem_. so i go back and work hard for a while ; either i push it a little bit further or i do nt but it s still there .
i ve got this residual set of problems that i hope to solve some day ; sometimes i do get them solved .
* nf : * you d worked on cart with several people , projection pursuit with john tukey and werner stuetzle , and ace with leo .
then what ?
* jf : * other stuff with werner , supersmoother ( friedman , @xcite ) and a paper on splines ( @xcite )
. then mars ( multivariate adaptive regression splines ) came after ace .
it started in the late 1980s .
i wanted a technique that would have the properties of cart except that it would make a continuous approximation .
one of the achilles heels of trees is that they make a discontinuous , piecewise constant approximation and that limits their accuracy . also , i d read de boor s little primer on splines ( @xcite ) which werner showed to me .
i d learnt most of what i knew about smoothing from werner .
smoothing was an important tool and i believe his thesis work had a lot about smoothing .
after that i knew something about splines , so i pieced together the idea .
you can think of cart as recursively making a spline approximation but with a zero - order spline which is piecewise constant , so i tried extending that so i could use a first - order spline which was a continuous approximation , discontinuous derivatives but a continuous approximation , and then you can generalize the approach to higher orders ( although in the implementation i did nt ) .
* nf : * as i recall , this ended up being a very large paper .
* jf : * yes , the mars paper was 60 pages of description and then there was another 80 pages of discussion , so it ended up as a 140-page paper ( @xcite ) . apart from mars
, i also developed a technique i called regularized discriminant analysis ( rda ; @xcite ) . some of my work was inspired by work that was going on in chemometrics .
there was a technique they called simca , which was basically a strange kind of quadratic discriminant analysis , viewed from a statistical perspective .
it s an acronym for soft independent modelling of class analogies ( @xcite ) . that was used a fair amount for classification problems in chemometrics .
* nf : * i recall a meeting involving some chemometricians where you and ildiko presented a paper on your views about pls , where you showed that it had some significant deficiencies .
have your views on this subject ever been accepted by the chemometrics people ? * jf : * i do nt think so , no .
i went to a chemometrics conference two or three years ago and everything was still pls after 20 years .
in the machine learning literature everything is a machine , every algorithm is called a machine . before that
, every algorithm was called a network in neural nets . in chemometrics
everything is called some kind of pls .
you reminded me : ildiko and i wrote a paper trying to explain pls from a statistical perspective ( @xcite ) .
also , when boosting came out much later , rob and trevor and i tried to show what it was doing , again from a statistical perspective .
we did pls and it turns out it s very close to ridge regression .
i do nt think the pls people appreciated it at all .
pls definitely has limitations .
one thing is that if the variables are all uncorrelated , then it does nt regularize at all .
at least ridge regression , which is very similar , still regularizes in that kind of situation .
so it depends upon the predictor variables being highly correlated to impose this regularization , whereas ridge regression , which gives pretty much the same result for highly correlated variables , also regularizes in the absence of a high degree of correlation .
* nf : * and rda ?
* jf : * rda related to this simca thing .
it was a very simple idea about linear discriminant analysis and quadratic discriminant analysis .
you consider an algorithm that is a mixture of the two .
then in the second part , when you do the quadratic discriminant analysis , you regularize the covariance matrices in a ridge style so there are two regularization parameters for the two covariance matrices , each being estimated separately .
each of the separate covariance estimates is blended with the common covariance , their average , with degree of blending being another parameter of the procedure .
i liked that idea .
we wrote the paper about pls when i was on my sabbatical in 1992 .
the sabbatical was broken up into small pieces , part of which was in australia .
that s when you and i started working on multivariate geochemical data .
* nf : * yes , that led us to prim .
would you like to say a little bit about prim ( patient rule induction method ) ?
* jf : * the idea there was hot - spot analysis .
data mining was coming in and one of the things that people wanted to do was look for needles in haystacks , hot - spots in data , for example , in fraud detection .
you expect a fairly weak signal , but what you hope for is that it s identified by a very sharp structure in at least a few of the variables .
prim ( @xcite ) was a recursive partitioning scheme but different from cart which was very greedy and aggressive .
that s where the `` patient '' comes in : it was meant to find a good split but only split a little bit and be patient and then look for another split , that was the idea .
* nf : * there was an earlier bias - variance paper in the 1990s .
* jf : * yes . there was kind of a cottage industry in the mid-1990s ; everyone was aware of the bias - variance decomposition of prediction error for squared error loss regression and it intrigued people to try and develop something analogous for classification . here
the loss is either zero or one , and the goal was a corresponding decomposition of the misclassification risk .
there were numerous papers on that .
leo wrote one ( @xcite ) and there were a lot in the machine learning literature .
i got the impression that you really could nt find such a decomposition , but what you could do was look at traditional bias and variance , which are well defined , and see how those two kinds of estimation errors , like bias and variance in estimating the probabilities , reflected themselves in misclassification risk .
so i wrote this paper ( @xcite ) where basically i showed that the curse of dimensionality affects classification much less severely than it does regression . in regression
, things get exponentially bad as the dimensionality increases , but not necessarily for many types of classification .
so that is why things like nearest neighbor and kernel methods , which do nt work terribly well in regression in high - dimensional settings , can perform reasonably well with classification : the curse of dimensionality does nt hurt them as much .
this is especially so with over - smoothing the density estimate : it can be very severe and can introduce huge error in the density estimate , but need not introduce much error in classification .
i did nt know where to publish that paper , or indeed whether to publish it all .
then a friend of mine , usama fayyad , contacted me .
he was one of the early people in data mining and may even have coined the term `` data mining .
'' he was starting a journal of data mining .
he said , `` i d like a paper from you in the first issue , '' so i said okay .
i had this one just sitting there , so i sent it off to him .
it turned out
and i did nt know this until much later that paper was read by a data mining fellow in israel , saharon rosset .
he felt that this showed that statistics could contribute to data mining .
so he decided he wanted to come to stanford and study .
he was one of the best students we ve ever had .
i learned a lot from saharon and still do .
so i would say the biggest success of that paper was that we got saharon to come to our department .
* nf : * when was it that you had the insight about high - dimensional data , that every point is an outlier in its own direction ?
* jf : * that came from projection pursuit . some time in the late 1980s , early 1990s , outlier detection was a big issue for people . i had seen it in various papers and talks .
i thought that it might be a natural application of projection pursuit .
projection pursuit looks for directions in the space such that when you project the data it has a particular interesting structure defined by a criterion that you then try to optimize .
so i thought , _
ok , we ll define a criterion that looks for outliers_. i came up with a criterion , programmed it up , tried it out and it was working beautifully .
it was finding all kinds of outliers and the nice thing about it is you see the projection .
so in that projection here is the data , here is the point , there is no other inference to be done ; it s an outlier , there it is .
i was very excited and after trying it on data , both simulated and real , i thought , _ well , we ve got to calibrate this .
how many outliers does it find when there are none ? _ i generated data from a multivariate normal distribution , tried the algorithm and it found this incredible outlier .
i thought , _
okay , that can happen , it s an accident _ , so i removed that point and searched again .
it found another one , another projection with a far outlying point .
it just kept doing this .
i could just peel the data .
i found this was very curious and i mentioned it to people and i believe it was iain johnstone who came up with the explanation that every point is an outlier in its own projection . that was the phenomenon that i was just discovering empirically .
* nf : * you ve mentioned the term `` data mining , '' which came from a nonstatistical community . what were your interactions with these other communities ?
* jf : * in the early 1990s i was becoming aware of the machine learning field .
i was invited to give a talk at a nips ( neuro information processing systems ) conference some time in the very early 1990s . that opened up a different world for me because there were all these people who were doing things with similar motivations but not with statistics , not in a statistical mode .
they were almost entirely algorithmically driven .
i felt that was wonderful , so i gave a talk there and i went back to those conferences throughout the 1990s .
* nf : * had they been aware of any of your work ? * jf : * well , they must have been aware of some of it because they invited me to give a talk .
i do nt know how much my work was referenced in their papers , probably some .
it was interesting the progression throughout the 1990s when i went to those conferences .
at the first one i attended there was lots of discussion of hardware and these were mostly electrical engineers .
in fact , there were two groups : the engineers who used neural nets and neural - net - type ideas to solve prediction problems ; and the psychologists who used them to try to understand the brain and how adaptive networks can learn things , the basic learning theory .
for the engineering part it is interesting how it evolved from a concentration on programs and hardware to looking more and more like statistics .
and now it s basically statistics .
they discovered bayesian methods .
i remember in early discussions with machine learning people i tried to explain why fitting a training data as closely as possible does nt necessarily give you the best future prediction , or what they call generalization error .
now they understand that completely , but in those days it was a little hard for some of them to grasp the concept .
to be fair , their interest was in very low noise problems like pattern recognition .
obviously there exists an algorithm that can tell a chair from a table every time , the brain can do it , so the bayes error rate is zero on that .
it s just that you ca nt come up with an algorithm to achieve the bayes error rate .
those were the kind of problems they were interested in .
so in that case fitting the training data as well as possible is the right strategy .
if the bayes error rate is zero , there s no noise .
* nf : * there were a number of distinct communities
* jf : * yes . there were three distinct fields , maybe more , that i know about .
there was statistics , there was artificial intelligence and then there was data base management .
* nf : * where did the computer scientists fit in ?
* jf : * computer scientists were doing data base management and artificial intelligence .
machine learning evolved , at least as far as i know , out of ai .
data mining originally emerged out of the data base management area .
it s all kind of a blend now and everyone is learning more of what the other people are doing .
the machine learners and data miners are learning more statistics and their research is looking more and more like statistics .
some statisticians are learning more about methodology and algorithms and their work is looking a lot more like machine learning or data mining .
* nf : * we ve talked about one or two students you were involved with before you joined the department , but once you joined you had some formal responsibilities to supervise these students . * jf : * yes , i had a number of students and i enjoyed them all in different ways .
one of the real advantages of being in an academic department is that you get to be around students with young fresh ideas and that eagerness that hasnt been stilted by time .
* nf : * what collaboration did you have with your students ?
* jf : * i certainly collaborated on their thesis work .
probably the student that i had the biggest and longest collaboration with was bogdan popescu , from romania .
* nf : * your style of doing things clearly influenced a lot of people who were around you at that time as students .
* jf : * i think so , yes . especially in my early days in the department my way of thinking about things was really very different ; it s not so much any more . we ve got rob and trevor and art [ owen ] , all of whom were students when i first came .
art was actually my student .
rob and trevor were not officially my students , but they came up to slac a lot . * nf : * they got infected by what they saw .
* jf : * trevor was werner s student , so he got infected strongly and so did rob , i think : the more phenomenological way of thinking , less the theorem proof theorem proof theorem proof approach . not that i devalue that approach .
i do nt want to give that impression , it s just different .
i m not good at it .
i do nt have the skill to do it .
* nf : * so far at stanford , we ve threaded our way through the 1990s and into the first decade of the new millennium and during this period you have commenced another very significant collaboration with some of your stanford colleagues .
* jf : * that s right . there were some very impressive and interesting developments in the machine learning field in the late 1990s and also in statistics .
one of them was leo s bagging idea , which was a very simple but clever idea .
then there were the boosting ideas that came out of the machine learning literature that were introduced by freund and shapire ( @xcite ) .
i started to become fascinated by this because it had a similar flavor to pls in the sense that it appeared to work reasonably well but it was nt clear why .
again , a cottage industry developed as to why .
the machine learners had their own approach using what they called the pac learning theory ( pac stands for probably almost correct ) , which was a way of looking at it which was very satisfying to them .
it was a good way to look at it , but i think we did nt quite understand it .
if it s analyzing data , it s doing what statistical algorithms do , therefore , there should be some sort of sound statistical basis for it .
so rob , trevor and i started a collaboration to try to figure out from a statistical point of view why this thing was working so well .
it was interesting in the sense that we did nt have the answer when we started the collaboration .
this was similar to working with leo , where we just posed the problem .
quite often when you form a collaboration you have an idea of the solution and you put it together , but we had no idea why this thing was working so well .
so we plodded along and got various insights along the way and i believe we figured it out ( @xcite ) , at least to our satisfaction but not to every one s satisfaction : leo never thought our explanation was the essential reason .
he thought our formal development was correct , but he did nt think that was the reason that led to boosting s apparent spectacular performance .
but i was convinced that we had explained it .
i think the machine learners , the pac learning people never thought so , i do nt think they completely understood the way we were looking at it .
* nf : * and since then you ve had an extremely productive collaboration with rob and trevor .
* jf : * yes . we did that in the late 1990s , then later in the mid-2000s i was asked to be an outside referee for a ph.d .
oral exam in the netherlands .
a student of jacq meulman s , anita vander kooij , was presenting her thesis and she had an idea . by this time the lasso which rob tibshirani had proposed in the mid-1990s was really coming on strong ; it still is .
@xmath3-regularized methods and the lasso , in particular , were really becoming popular .
there was a cottage industry on developing fast algorithms for doing it .
engineers had worked on this , machine learners had worked on this , and there was a spectacular paper by brad efron and some colleagues ( @xcite ) .
so this was very active at the time .
then anita and jacqueline had this really simple idea that a professional on optimization would dismiss out of hand , namely , just doing it one at a time .
they were working on a computer program that involved optimal transformations of the variables , and for this they were using the back - fitting algorithm . including regularization
then turned out to be simple .
lots of people had developed the idea of optimizing one at a time .
this is usually dismissed in optimization theory as not performing well which is correct unless the one - at - a - time solution can be obtained very conveniently and rapidly : then it can become competitive .
werner and i had explored this with our so - called back - fitting algorithm in projection pursuit regression and to fit additive models as well . anyway , their idea was that you hold all of the coefficients fixed but one and then solve for the optimal solution for that one .
this can be done very fast .
then you just cycle through them .
they developed it independently , but it was not a new idea : other people had developed it before , but it did nt seem to have been taken very seriously .
so when i came back from the netherlands i told rob and trevor about this and they got excited and we started working on applying the idea to a wide variety of constrained and regularized problems and continue to do so to this day .
rob and trevor and their students have come up with all kinds of new regularization methods , how we can do things one at a time and make it go very fast .
we applied it to the lasso and to the elastic net , which was something that trevor and a student , hui zou , had done in the mid-2000s ( @xcite ) .
it s a continuum of regularization methods between ridge regression and the lasso .
you dial in how much variable selection you want . in ridge there s no variable selection , lasso does moderate variable selection , so we extended it to the elastic net .
jacqueline and anita had also extended it to the elastic net .
then we extended it to other glms , logistic regression , binomial , poisson , cox proportional hazards model , and put together a whole package called _ glmnet _ that seems to be widely used now .
it allows you to do all these different regularized regressions with the various different glm likelihoods , and that work is still going .
i like writing the programs because they seem to run faster than other people s .
it is probably because of my impoverished youth when i worked on computers that were nothing like the computers now and you really had to write efficient programs .
that skill seems to have remained with me .
* nf : * this collaboration with rob and trevor resulted in a particularly important publication .
* jf : * yes , our book ( @xcite ) . that turned out to be an unbelievable success and i helped with parts of it , but it was mostly written by rob and trevor .
it just hit the right niche at the right time and i guess it is still selling very well , but you can download a pdf version from the web for free now .
* nf : * just to pick up on the point you made about you doing the programming , i remember you told me years ago that you had nt solved the problem until you d written the code to demonstrate the technique .
* jf : * i do nt have the requisite skills to do all the theory .
the only way i can see if it s a good idea is if i program it up and try it out , test it in a wide variety of situations and see how well it works .
* nf : * let s pick up some parallel activities that you d been engaged in , starting with mart . * jf : * at the time of my second lengthy visit to australia in 1998/1999 , i was fascinated with the boosting idea .
mart was a kind of a spin - off from the work that i d done with rob and trevor on trying to understand how boosting works .
i got a few ideas for how to extend boosting . boosting
was originally developed as a binary classification problem and while i was visiting csiro in sydney i wanted to extend it to regression and to other kinds of loss functions , so i developed this notion of gradient boosting which evolved into what i called mart , multiple additive regression trees .
i wrote that program ( also called mart ) and developed those ideas .
that was my rietz lecture i believe , which was published in the _ annals _ ( @xcite ) , an unusual paper for them to publish .
i still wanted to understand more about why boosting was working .
one of the ideas that i had developed with the gradient boosting was the idea again a sort of a patience idea that you can think of boosting as just ordinary stepwise or stage - wise regression .
you fit a model , say , a tree ( most people use trees ) , you take the residuals and then you fit a model to the residuals . you take the residuals from the sum of those two trees and build another model based on those residuals . now that s very greedy ; every time you re trying to explain as much about the current residuals as you can with the next model .
i came up with an idea ( again back to patient rule induction ! ) that when one finds the tree that best fits the residuals , only add a little bit of that tree , in other words , shrink its contribution .
so you multiply that tree by a small number like 0.1 or 0.01 , before it s added to the model . that turned out to really improve the performance .
so i wanted to understand why it was improving the performance and try to understand more about gradient boosting .
this was the time when bogdan popescu was my student .
he showed that the shrinkage only affected the variance and not the bias .
i thought this was a very important clue .
then , along with other people , we found that what this was doing was a kind of lasso . if you did nt do the shrinkage , then you were doing something like stepwise or stage - wise regression .
it produced solutions that would be very similar to the lasso and if you followed that strategy in a linear regression , it produced solution paths very close to the lasso .
in the beginning we thought they might be identical because we ran a few examples and they produced identical paths .
it turns out that it will only produce identical paths in two dimensions or if the lasso paths are monotone functions of the regularization parameter .
saharon rosset did nice work in this area , as did others .
there was part of an issue in the _ annals _ devoted to boosting [ _ annals of statistics _ * 32*(1 ) , 2004 ] .
there were several very heavyweight theoretical papers , very fine papers , showing that connection , showing that boosting was consistent provided you regularized in this way .
* nf : * jerry , you ve had long - term enthusiasm for acronyms .
what do isle and rulefit stand for ?
* jf : * isle stands for importance sample learning ensembles . again , throughout this time i was interested in why the ensemble learning approach was so effective and isle was a different way of looking at ensemble methods .
the idea was that you define a class of functions , and pick functions from that class .
the first thing that occurred to me was that with boosting and bagging and other ensemble methods you just kept adding trees .
there were some people who thought , _
okay , if you have an ensemble , how do you figure out what is the optimal way to weight each tree ?
_ i thought that was a very simple problem : if i want to have a function that s linear in a set of things , i know how to find the coefficients , that s called regression . at this time , leo was doing random forests and a lot of people were doing boosting .
i suggested that once you get the ensemble , you just do a regularized regression to get the weights of each of the trees or whatever they may be .
each element of the ensemble in machine learning literature is called a base learner , or a weak learner , because generally no one of them by themselves is very good , but the ensemble of them is very good .
that was one of the things that we understood about why boosting worked .
one of the reasons why boosting was so surprising was that in machine learning literature they had a notion of weak learners and strong learners : a weak learner is one that has low learning capacity and a strong one has high . there was a lot of impressive theoretical work by rob schapire , who was one of the co - inventors of the original successful boosting algorithm .
he showed that with this boosting technique you could take a weak learner and turn it into a strong learner , as long as the weak learner could achieve an error rate of @xmath4 above 50% .
this was very lovely work .
but when you deal with it from this linear regression perspective it does nt seem so surprising .
we ve encountered many problems where just one variable alone ca nt do much , but a number of variables fitted together in a regression can do very well . from my statistical perspective that s what s happening .
i thought you could do this with a lot of different things .
if you have a class of functions , you pick functions from this class and then you do a linear fit . then the question is : how do you pick the functions from the class ? if you just randomly pick them , nearly all of the functions will have no explanatory power as will their ensemble . if you pick them to all
be very strong , then their outputs are all highly correlated and you are not gaining anything from the ensemble
. the ensemble will give the same predictions as any one of one of them .
so you have a trade - off that leo had discussed a lot .
you do nt want your learners in the ensemble to be highly correlated in their predictions , but you do want them to have some predictive strength . that s a trade - off .
this was well known before the post - fitting idea .
the lasso and other regularization methods are natural for the post - fitting because they could be applied even when the size of the ensemble is much larger that the number of observations .
so you needed fast algorithms for the lasso and other regularized regressions that were being developed around that time .
it was a convergence of things .
rulefit was an ensemble method totally motivated by this concept .
the main difference was that instead of fitting an ensemble with boosted trees and then doing the post regression , you would take the trees , decompose them into rules , forget the trees the rules came from , and use them as a batch of `` variables '' in a linear fit .
leo made a remark once , maybe in the mid-2000s shortly before his death , that the real challenge in machine learning is not better algorithms , grinding out a tiny bit more predictive accuracy .
our very best learning machines tend to be black - box models neural networks , support vector machines , ensembles of decision trees and they have very little if any interpretive value .
they may predict very well , but there is no way you can tell your client why or how it is making a prediction , why it made that prediction rather than another one .
he thought that the real challenge was interpretability and he had put some interpretational tools into his random forest , namely , the relative importance of the predictor variables and some other things .
i wanted to see if there was some way to do interpretability and the idea was that if you have an ensemble method , it s basically a linear model and linear models are very interpretable as long as you can interpret the constituents , the actual terms in the model .
trees you can interpret , but i thought that it s easier to interpret rules .
a tree produces a rule derived from the path from the root to a terminal node : that s why it s so interpretable .
it can tell you exactly what variables are used to make the prediction and how it used them , which is why trees are so popular .
rule - based learning has also been a real staple in machine learning throughout its history .
so i thought of breaking up the tree into its rules , putting the rules together in a big pot and then doing a lasso linear regression on the rules .
the hope was that since the rules are nt very complicated and are easy to interpret , you could make much more interpretable models .
that in and of itself was only partially successful .
but along the way i developed ways for assessing the importance of the variables for individual ensemble predictions .
another thing that i did in that work was to develop some techniques for detecting interaction effects , seeing what variables were interacting , exploring interaction patterns of the variables .
so that was rulefit .
i havent done much beyond that in developing general learning machines like mars and mart , etc .
rulefit is my last one so far .
* nf : * i dare say there will be more to come .
you ve been in the stanford department of statistics now for over thirty years .
how have you found it as an environment for a statistical scientist ? * jf : * unbelievably great , i ca nt think of a place i d rather be .
my greatest joy is to have an office in the hall with so many bright and famous people .
my nearest office neighbors are brad efron , percy diaconis and wing wong , along with all the other fantastic people down the hall .
it s such a stimulating environment .
everyone is so sharp , so smart , so inventive and original .
you take it for granted after a while , but when you visit other places you find it s not like that everywhere .
i consider it great good fortune that i was able to join that department and i thank them for accepting me , because i was a kind of an odd appointment at the time .
* nf : * i am sure you look like a mainstream appointment right now .
do you feel that the department has , to some extent , progressed towards you ?
* jf : * okay , maybe a little bit , yes .
* nf : * what are your current interests ?
* jf : * there s the whole regularization idea which i still think is fascinating
. there are some leftover questions that current research has not yet answered .
i d like to think more about that area .
another area is improving decision trees .
trees have emerged as being very important largely , in my view , because of the ensemble methods .
trees have very nice robustness properties .
they can be built quickly , they are invariant to monotone transformations of the predictors , they are immune to outliers in the predictors , they have elegant ways of handling missing values and of incorporating both numeric and categorical variables .
they are a very nice type of learning machine : you just pour the data in , and you do nt have to massage the data too much prior to that .
they have several achilles heels , one of which was of course accuracy , but i think that s been solved by the ensemble methods , which carry over all these advantages while dramatically improving their accuracy : not just by 10% or 20% , but sometimes by factors of 3 or 4 .
i think boosting is one of the key ideas of machine learning .
it has really advanced both theory and practice .
another achilles heel of trees is categorical variables with a very large number of levels .
back when we were doing cart , a typical categorical variable might have 6 levels . now it s routine to have hundreds or thousands of levels .
that destroys trees because there is no order relation .
the number of possible splits grows exponentially with the number of levels .
optimizing over all these possibilities can lead to severe over - fitting . in situations
where there s a substantial amount of noise , this can lead to spurious splits that mask the truly important ones .
so that s left over and it is one of those things that i mentioned i keep in the back of my mind and every so often try to think about again , which is what i m doing now with this one .
another thing i have been thinking about recently is the issue that many of the problems arising with data that is seen now , especially commercial data , tend to be binary classification problems . in my industrial consulting
i see much more classification than regression .
this is surprising because historically most statistics research has centered around regression .
classification was something of a back issue in statistics . in machine learning
, classification has always been the main focus .
in fact , they refer to regression as classification with a continuous class label .
a lot of the data is highly unbalanced , you may have millions of observations but one class has very few . in engineering and machine
learning they tend to label the class as @xmath5 and @xmath6 .
usually there is a very small fraction of positives , like in fraud detection , for example , where you have a data base with a huge amount of data , but the number of instances of fraud is a small fraction of the data at least you hope that s the case !
it s certainly true in e - commerce , where the rate of clicking an ad on a page is around 1% and then the conversion rate ( which means you click the ad and then and go buy something ) is two orders of magnitude lower than that .
so the issue is how to deal with data like that and there are rules of thumb that say if you have , say , a hundred positive examples in a million negatives , you do nt use all the million , you randomly sample them .
so then the question is : _
what s the strategy and how many do you need ? _ and there s another rule of thumb that says if you have 5 times as many negatives as positives , that s all you really need .
i doubt that s true in general , but i d like to be more precise about it because it s of huge practical importance : if you have millions of observations which you can randomly sample down to a thousand or a few thousand , that totally changes the dynamic of how you do your analysis .
so that s another thing i m thinking about .
trevor hastie and a student , will fithian , recently did some nice work ( @xcite ) in this area in the context of logistic regression .
another area of current interest is loss functions .
a machine learning procedure is specified by a loss function on the outcome and a regularization function on the model parameters .
defining appropriate regularization functions and their corresponding estimators for different problems is currently a hot topic for research in machine learning and statistics .
there is an avalanche of papers on the subject .
there seems to be less interest in finding appropriate loss functions for different problems .
the loss function @xmath7 specifies the loss or cost when the true value is @xmath8 and the model predicts @xmath9 .
i have found in my consulting work that being able to customize the loss function for the problem at hand can often lead to big performance gains .
most applications simply use the defaults of squared - error loss for regression and bernoulli log - likelihood on the logistic scale for classification .
i d like to investigate broader classes of loss functions appropriate for certain kinds of specialized problems that go beyond the ones usually used in glms .
i find i spend a lot of my time on my programs .
i put most of my programs on the web and people can download them and use them , and they report bugs back and i feel obligated to try to fix them .
as you go along in your career and you ve done more and more things , you have to spend more and more of your time back - caring feeding those things as well as moving forward .
i ve had a long career now and spend a non - negligible amount of my time just maintaining past stuff .
* nf : * it s like entropy , is nt it , always increasing .
the list of errata never shrinks .
* jf : * yes . then people have questions , they do nt understand things , or people use the algorithms in ways that you never dreamt they might be used .
something else i ve just thought about . in the mid-1990s i worked a lot on trying to incorporate regularization with nonconvex penalties .
i spent a fair amount of time on a technique which is somewhat similar to the boosting technique but in the linear regression context .
the lasso imposes moderate sparsity as opposed to an @xmath10-penalty ( all subsets regression ) which induces the sparsest solutions .
so i did a lot of work spanning the gap between all subsets which is very aggressive variable selection and which often does nt work , especially in low - signal settings and the lasso , which is moderately aggressive in selecting the variables . that involves nonconvex penalties .
the lasso is the sparsest inducing convex penalty .
of course with convex penalties , as long as you have a convex loss function , then you have a convex optimization which is a lot nicer than nonconvex optimization when you have multiple local minima and other problems .
so i did spend a lot of time working on boosting techniques applied to linear regression with nonconvex penalties .
* nf : * statisticians around the world have been using your techniques for a long time now , there s a company that exists simply to sell your software and you generated that industry .
also your ideas and methods were used by yahoo!. * jf : * yes . they used the commercial analog of mart as a big part of their search engine .
i do nt know exactly what they use now , maybe the microsoft search engine .
but for a long time , mart was an integral part of the yahoo ! search engine .
* nf : * let s actually leave statistics briefly , because you do have a life outside statistics .
* jf : * well , somewhat .
( see the anecdote _ life outside statistics _ in @xcite . ) * nf : * and then there s been your long - time interest in gambling and computers .
* jf : * yes , that started when i was a graduate student .
( see the anecdote _ statistics , computers and gambling _ in @xcite . )
* nf : * finally , let s step back , or maybe move to a greater height in this conversation , in the sense of taking a perspective on statistics at certain times . there have been at least two occasions ( friedman , @xcite , @xcite ) when you have committed your thoughts to print about `` where are we now with statistics and computing ? ''
let s go back to 1987 when there was a symposium on `` statistics in science , industry and public policy . ''
you were invited to present a paper on `` modern statistics and the computer revolution . ''
* jf : * it was an assignment i could nt refuse because it was from the person in charge of statistics funding at nsf .
i had an nsf grant at the time , so i had to go back and give a talk about what i thought about the future of statistics and how computing might affect statistics in the future . *
nf : * in this paper you talked about automatic data acquisition , some of its benefits and also some of the issues that it raised . early in the paper you said that `` what separates statistics to a large degree from the information sciences
is that we seek to understand the limits of the validity of the inference .
'' do you think that separation is still the case in , say , machine learning areas ?
* jf : * not as much as it was , but i would say so .
the huge contribution of statistics to data analysis is inference , what you are getting out of the data , or learning from the data .
how much of it is really valid . that has been the main thrust of statistics .
it has become less of a thrust only because data sets have gotten larger and so the sampling variation has become less of a problem , but it s still there in a big way .
originally , i think there were people in neural networks and machine learning who were nt very concerned about that at all , whatever they found they assumed was reality . and to be fair , at least in machine learning , that was because they were dealing with pattern recognition problems where the inherent noise was not large , where the bayes error rate in a classification problem was really very close to zero if not zero .
the particular classifier that attained that error rate was complicated and hard to get at .
so i do nt think that inference was as big a problem in those kinds of things .
statisticians originally came from other areas where the data sets were small and signal to noise was very low . in
those settings inference is a very important part of the learning procedure .
* nf : * but the computer scientists and the machine learners havent stayed in their little box , they started playing with other problems .
* jf : * oh yes , bayes - type ideas are now spreadthroughout machine learning , computer science and engineering , for example .
inference is there , although it s perhaps not given quite the high priority that we statisticians give it .
* nf : * you commented that most of the methods being used in statistics in 1986 were actually developed before 1950 , but that the computer was liberating us from these mathematical bindings such as closed - form solutions and unverifiable assumptions .
i particularly like your closing comment that `` the cost of computation is ever decreasing but the price we pay for incorrect assumptions is still staying the same . ''
would you care to amend that statement now ?
* jf : * no , i think it s the same , we have to make fewer and fewer unverifiable assumptions these days .
the sample reuse techniques like cross - validation and the bootstrap have really freed us up ; they have really helped the kind of thing i do a lot .
quite often when you come up with a new complicated procedure and someone will say , `` how do you do the inference ?
how do you put error bars in ? '' or something like that , you just reply , `` well you can bootstrap it . ''
so that was a giant contribution to statistics .
but in the area that i work in it is especially valuable .
* nf : * moving on 12 years , you had another opportunity to take a helicopter view at the isi meeting in helsinki , where there was a session on `` critical issues for statistics in the next two decades . ''
you presented a paper on `` the role of statistics in the data revolution ? , '' and i note the question mark at the end of that statement ! in the summary you said , `` the nature of data is rapidly changing .
data sets are becoming increasingly large and complex .
modern methodologies for analysing these new types of data are emerging from the fields of data base management , artificial intelligence , machine learning , pattern recognition , and data visualization .
so far , statistics as a field has played a minor role .
this paper explores some of the reasons for this and why statisticians should have an interest '' and so on .
what i m interested in is : how have things changed since then , what needs to be done , and what s blocking this change ?
* jf : * oh , i think it s changing quite a bit .
perhaps i have a nonrepresentative view being at stanford , but i think that statistics is definitely moving forward in those areas .
statistical research in data analysis is definitely overlapping more with machine learning and pattern recognition . as i pointed out in the 1987 paper , and as i say whenever i am asked about the future of statistics , you ca nt answer that question , you have to ask : _ what is the future of data ? _ statistics and all of the data sciences will respond to whatever data is present .
no one could have anticipated gene expression arrays in the late 1980s .
now statisticians have adapted to that and the whole bioinformatics revolution as well , making huge contributions to those areas .
* nf : * in particular , in the 1999 paper , reflecting on the relationship between statistics and data mining , you said that `` from the perspective of statistical data analysis , however , one can ask whether data mining methodology is an intellectual discipline .
so far the answer is : not yet ''
has the answer changed or has the question become irrelevant ? * jf : * that s a good question .
i would say it s relevant and changing ; i m not sure that it has totally changed yet .
i think you need people who can come up with several ways of looking at data but who perhaps do nt have the requisite skills to understand at a basic level what s happening .
and you need people who are very skilled at taking a methodology and a situation and then deriving the properties of the method in that situation .
i think the attitude in the data mining community is : `` if it works , great ! we ll try things and we ll find out the things that work . ''
i think that s a perfectly reasonable way to proceed . some people like to proceed from basic principles : let s first understand the basic principles , and from there develop the right things to do , or good things to do .
the other is an _
ad hoc _ approach just think hard about the problem , try to figure it out which is the way tukey did it way back when and try to come up with something that works well .
that approach is fraught with danger , of course : not everyone is as smart as tukey .
as people develop techniques , they advertise and convince people they are really very good when they are not , so one has to be careful . but generally , if there s a methodology like pls , support vector machines , boosting or more general ensemble methods that seems to repeatedly work very well , there s probably a good statistical reason , even if in the beginning it was not known .
the understanding , the underlying principles of why they work well , came later on .
* nf : * later in that paper you said , `` perhaps more than any other time in the past statistics is at a crossroads ; we can decide to accommodate or resist change . ''
have we accommodated , are we still resisting change , how do you situate statistics now in the information sciences ? * jf : * i think statistics is accommodating change , not as fast as i would like , faster than some other people would like , but certainly adapting to change .
ultimately it is data that s driving statistics as well as the other information sciences .
but i think statistics today is much more responsive .
when new forms of data come out there are statisticians who immediately see the opportunity , as well as engineers and other people .
* nf : * then in a sense i think you have answered your concluding remark in this paper , which was : `` over the years this discussion has been driven mainly by two leading visionaries of our field .
john tukey in his 1962 annals of mathematical statistics paper ( @xcite ) and leo breiman at the 1977 dallas conference . over twenty years have passed since that conference .
we again have the opportunity to re - examine our place among the information sciences . ''
so you feel that we are sitting rather more comfortably in there than we did ?
* jf : * again , being at a wonderful place like stanford , i think so , yes . i think we are doing it right .
we havent abandoned our tradition of formal inference , which is very good because that s something that the other information sciences do nt do nearly as well as we do .
there are isolated incidences of people in those other areas who do it very well , but it s not the priority that it is in statistics .
that being one of our priorities really helps a lot because you must understand the limits of inference at some point .
i think in the early stages we were trying things out , seeing what would work and using our intuition and i think the insights are beginning to come .
these days , if you look at the work being done in bioinformatics , in computer science and in statistics , there s a huge overlap where there was nt before in attitude as well as the actual work , the problems we re trying to solve .
* nf : * well , we still hold true to the guiding standard of understanding and managing variability .
* jf : * that s right , and i think we pay more attention to that than other fields do and i think that s good . in the past , perhaps we may have paid too it much attention .
well , not _
too _ much attention because statistics was working on methodology for a certain kind of data small data sets , high noise , where inference was everything , are you seeing a signal or not ?
this is the essence of hypothesis testing .
not _ how big is the signal and what are its properties ?
_ , just _ can we say whether there is one or not ?
_ with those small data sets and high noise often that was the only thing you could ask .
hypothesis testing was a huge intellectual triumph .
but now with larger data sets and better signal - to - noise ratios , we can start asking more detailed questions : _ what is the nature of the signal ?
what variables are participating in the prediction problem ?
how are they participating ?
how are they working together to produce the result ? _ * nf : * you have certainly changed
the way a lot of people think about statistics and you believe you are doing statistics and have done consistently . if colin mallows were in our presence now , do you think he would be describing what you do as statistics ?
* jf : * i guess you d have to ask him . perhaps .
i have always believed that
perhaps erroneously!but i always believed that i was doing statistics .
you know what they say : a rose by any other name would smell as sweet .
i think there is less of a need to categorize things . who cares about the name of what you re doing as long as it s interesting and potentially useful .
the categories seem to be all blurred now and that s all to the good . *
nf : * okay , well by a miracle of modern science we have sitting beside us a reincarnation of jerry friedman , except he s only twenty years old , and he s wondering what to do at college .
what are you going to recommend ?
* jf : * what i always recommend whenever i m asked : `` what should i study , what should i do ? ''
i always say , `` study and follow what you are most passionately interested in
. do nt worry about what skills are going to be marketable in ten years because that will all change . ''
if you go to school to learn a skill that you do nt like because you think it is going to be especially marketable when you get out 5 or 6 or 8 years from now , that could change .
you ve suffered through all of that and you end up without marketable skills after all . at least if you study something you re really enjoying or are passionate about , you ve had all that fun .
if you re lucky like i was and it turns out that your skill evolves into being marketable , then so much the better . follow your passion .
* nf : * you think statistics might easily be one of those ? * jf : * oh , i agree with hal varian ( chief economist , google ) , who made that statement , that statistics is going to be the glamor field of the future for some time ( `` i keep saying the sexy job in the next ten years will be statisticians . ''
people think i m joking , but who wouldve guessed that computer engineers wouldve been the sexy job of the 1990s ?
the data revolution using data to answer questions and solve problems has really emerged .
not so long ago when , say , you were at a factory or at some kind of production line and yield was going down , what did you do about it ?
well , you called on the supervisors and experts , you got into a room and you tried to figure why yield might be going down . it did nt often occur to people to collect data .
now everybody collects data .
almost every production line and factory is heavily instrumented at every point and data is being collected .
in fact , i think maybe it may come to the point where people ask too much of data ; data ca nt answer every question .
* nf : * i was pondering how to title this conversation and i did have in mind something like `` jerry s search for pattern , '' but then it occurred to me that a pattern is only a pattern *
jf : * but a good cigar is a smoke .
i agree .
* nf : * somebody once said something along those lines .
* jf : * yes , it was kipling of course ( e.g. , @xcite ) .
i started smoking cigars on and off when i was young , in high school and just out of high school .
i worked for the forestry service fighting forest fires and surveying timber access roads .
where i lived most of the countryside was national forest and so that was a traditional job to do . at one of the camps
the only facilities were out - houses and they smelled very , very bad .
it was a real ordeal to use them , especially if you had to stay longer than ten or twenty seconds .
the only way that i could stand to do it was to light up a really foul - smelling cigar , and smoke it while i was in there .
that s why i started smoking cigars .
i smoke better cigars now .
* nf : * so do you feel we should stop talking about patterns right now and adjourn ?
* jf : * it would nt be a bad idea .
* nf : * well then , many thanks , jerry , for this glimpse of a fascinating scientific odyssey .
i feel as if i ve been slip - streaming slim pickens , riding a rocket down the years in which statistics and computing have become inextricably intertwined , except you ve been sitting on the nose - cone and pointing the rocket , which slim pickens did nt quite have the ability to do .
may you ride for a long time to come .
* jf : * well , thank you very much , nick , i really appreciated it .
the author thanks rudy beran and bill van zwet for valuable critical comment on a draft of this article , the editor and an associate editor for their helpful feedback , and jerry for his hospitality during the interview and patience during the preparation of the article .
the work was supported in part by valuemetrics australia .
the supplementary materials associated with this article comprise a number of anecdotes , plus an example of one way in which john tukey communicated his research ideas to jerry in the course of their collaboration .
they are available from @xcite . | jerome h. friedman was born in yreka , california , usa , on december 29 , 1939 .
he received his high school education at yreka high school , then spent two years at chico state college before transferring to the university of california at berkeley in 1959 .
he completed an undergraduate degree in physics in 1962 and a ph.d . in high - energy particle physics in 1968 and was a post - doctoral research physicist at the lawrence berkeley laboratory during 19681972 .
in 1972 , he moved to stanford linear accelerator center ( slac ) as head of the computation research group , retaining this position until 2006 . in 1981 , he was appointed half time as professor in the department of statistics , stanford university , remaining half time with his slac appointment .
he has held visiting appointments at csiro in sydney , cern and the department of statistics at berkeley , and has had a very active career as a commercial consultant .
jerry became professor emeritus in the department of statistics in 2007 .
apart from some 30 publications in high - energy physics early in his career , jerry has published over 70 research articles and books in statistics and computer science , including co - authoring the pioneering books _ classification and regression trees _ and _ the elements of statistical learning_. many of his publications have hundreds if not thousands of citations ( e.g. , the cart book has over 21,000 ) .
much of his software is incorporated in commercial products , including at least one popular search engine .
many of his methods and algorithms are essential inclusions in modern statistical and data mining packages .
honors include the following : the rietz lecture ( 1999 ) and the wald lectures ( 2009 ) ; election to the american academy of arts and sciences ( 2005 ) and the us national academy of sciences ( 2010 ) ; a fellow of the american statistical association ; paper of the year ( _ jasa _ 1980 , 1985 ; _ technometrics _ 1998 , 1992 ) ; statistician of the year ( asa , chicago chapter , 1999 ) ; acm data mining lifetime innovation award ( 2002 ) , emanuel & carol parzen award for statistical innovation ( 2004 ) ; noether senior lecturer ( american statistical association , 2010 ) ; and the ieee computer society data mining research contribution award ( 2012 ) .
the interview was recorded at his home in palo alto , california during 34 august 2012 .
./style / arxiv - general.cfg |
in many astrophysical settings , emission of radiation via rotational and vibrational transitions of molecules plays an important role in cooling warm gas .
because the emission of such radiation is in sharp spectral lines , this mechanism is referred to as line cooling .
line radiation from the water molecule @xmath2 , with its permanent electric dipole and its high cosmochemical abundance , is significant in a variety of settings ranging from molecular clouds to protostellar envelopes ( e.g. , cernicharo & crovisier 2005 ) .
more recently , line cooling from @xmath2 molecules has been recognized to play a pivotal role in the energetics following the passage of a shock wave through the dense gas in the solar nebula protoplanetary disk ( see desch et al .
2005 ) .
the rate of emission of line radiation from a warm gas containing water molecules is difficult to calculate , because there are so many accessible rotational and vibrational energy levels , and therefore a great many transitions . in dense molecular gas ( @xmath3 ) , these energy levels are populated according to boltzmann statistics ( i.e. , in local thermodynamic equilibrium , or lte ) , but at lower densities the populations must be calculated by balancing transition rates .
finally , the cooling of gas by line radiation relies on the ability of the photons to escape the system without being reabsorbed ; to the extent that other , nearby molecules can absorb the emitted photons , the gas does not cool .
photons generated during a molecular transition are emitted and also reabsorbed over a small range of frequencies centered on the line frequency , with an efficiency that depends on the frequency shift from the line center .
rather than integrating the equations of radiative transfer over all frequencies within the line , all standard treatments of the problem instead assume integration over this line profile " and use a frequency - integrated escape probability of photons .
this escape probability is a function of the column density of the molecule within the system .
this is the approach taken by neufeld & kaufman ( 1993 ; hereafter nk93 ) , in particular .
the calculation of nk93 of the rate of line cooling from @xmath2 molecules has remained the state of the art for 15 years ; but , as we discuss below , several aspects of the nk93 calculation are now out of date .
their calculation uses a limited database of transitions , an oversimplified escape probability formulism , and ignores absorption by dust grains .
it is one of the goals of the work here to update the @xmath2 line cooling rates to improve on the calculations of nk93 .
a second goal of this paper is assess whether or not line cooling plays a significant role in cooling the gas following the passage of a shock through the dense gas of the solar nebula . as discussed by desch et al .
( 2005 ) , previous modeling has not determined whether or not line cooling can be neglected .
this has significant implications for the issue of chondrule formation . the parent bodies of the most primitive meteorites , the chondrites , formed at about 2 - 3 astronomical units ( au ) from the sun , 4.57 billion years ago ( wadhwa & russell 2000 ) .
chondrites are the most primitive meteorites in our collection , in that they have suffered very little alteration since their formation , and therefore contain information about the conditions that existed in the early solar nebula .
chondrites are remarkable for containing calcium - rich , aluminum - rich inclusions ( cais ) , the oldest solids in the solar system , whose formation has been dated to @xmath4 mya ( amelin et al .
2002 ) . also found in abundance within chondrites are sub - millimeter- to millimeter - sized , ( mostly ferromagnesian ) igneous spheres , called chondrules , from which the chondrites derive their name .
chondrules formed , at most , @xmath5 2 - 3 million years after cais ( amelin et al . 2002 ; kita et al .
2005 ) , as melt droplets that were heated to high temperatures while they were independent , free - floating objects in the early solar nebula .
after they were heated , cooled and recrystallized , chondrules were incorporated into the parent bodies from which chondrites originate .
chondrules are capable of providing incredibly detailed information about conditions in the solar system protoplanetary disk , if the process that led to their heating , melting and recrystallization could be understood .
chondrules make up to 80% of the volume of ordinary chondrites ( grossman 1988 ) , and it is estimated that @xmath5 10@xmath6 g of chondrules exist in the asteroid belt today ( levy 1988 ) .
such a prevalence of chondrules suggests that chondrule - forming events were widespread in the solar nebula . a process that can melt @xmath7 of rock is surely a dominant process in the solar nebula disk .
the chondrule formation process , despite its obvious importance , has been a mystery in the field of meteoritics for two centuries ( sorby 1877 ) .
proposed mechanisms include interaction with the early active sun , through jets ( liffman & brown 1995 ; liffman & brown 1996 ) or solar flares ( shu et al . 1996 , 1997 , 2001 ) , melting by lightning ( pilipp et al . 1998 ;
desch & cuzzi 2000 ) , and melting by planetesimal impacts ( urey & craig 1953 ; urey 1967 ; sanders 1996 ; lugmair & shukolyukov 2001 ) . the dominant theory , though
, is that chondrules were melted in shock waves in the protoplanetary disk ( hewins 1997 ; jones et al .
2000 ; connolly & desch 2004 ; connolly et al .
a shock wave , or shock front , is a sharp discontinuity between supersonic and subsonic gas , over an area only a few molecular mean free paths thick , typically only meters in the solar nebula .
the gas is slowed , compressed , and heated by the time it reaches the other side of the shock front .
solids moving with the gas are heated not only by thermal exchange upon entering the shocked region , but also by friction as they are slowed to the post - shock gas speed , and by absorbing the infrared radiation emitted by other solids .
the shock model of chondrule formation appears able to resolve the chondrule formation mystery , because it makes several detailed predictions about chondrule formation that are largely borne out by observation and experimentation , especially regarding chondrule thermal histories .
heating and cooling rates of chondrules have been determined experimentally . according to furnace experiments , in which melt droplets with chondrule compositions
are allowed to cool and crystallize , reproduction of chondrule textures requires particular ranges of cooling rates between the liquidus temperature ( @xmath8 k ) and solidus temperature ( @xmath9 k ; hewins & connolly 1996 ) .
porphyritic chondrules cooled at about @xmath10 , and barred - olivine chondrules cooled at about @xmath11 ( hewins et al.2005 ; see also desch & connolly 2002 ) .
additionally , chondrules retain volatile elements such as s , indicating that they did not remain above the liquidus for more than minutes , and cooled quite rapidly ( @xmath12 while above the liquidus ( yu & hewins 1998 ) ) . finally , there is no indication of the isotopic fractionation that would arise from the free evaporation of alkalis such as na , which constrains the time spent at high temperature before melting ( tachibana et al .
modeling of isotopic fractionation has shown that chondrules must heat up from 1300 to 1600 k on the order of minutes or less ( tachibana & huss 2005 ) . passage through nebular shocks
satisfies nearly all the experimental constraints on chondrule formation in broad brush ( iida et al .
2001 , hereafter insn ; desch & connolly 2002 , hereafter dc02 ; ciesla & hood 2002 , hereafter ch02 ; desch et al .
prior to passage through the shock front , chondrules are moderately heated by absorbing radiation emitted by chondrules which have already passed through the shock front . upon passage through the shock front ,
the gas is immediately slowed , and compressed and heated , but the chondrules continue at supersonic speeds through the gas .
they achieve their peak heating during this stage , due to absorption of radiation , thermal exchange with the gas , and supersonic frictional drag heating .
this stage lasts until the chondrules slow to the gas velocity , which takes an aerodynamic stopping time , @xmath13 minute ( where @xmath14 and @xmath15 are the particle density and radius , and @xmath16 and @xmath17 are the post - shock gas density and sound speed ) . at this time , the gas and chondrules are dynamically coupled and chondrules are heated only by absorption of radiation and thermal exchange with the gas .
soon thereafter the solids and gas become thermally coupled as well , and both components achieve similar temperatures .
although the chondrules pass through the first 100 km of the post - shock region more rapidly than the gas , once they do , they will thermally equilibrate to the gas temperature .
once this happens , the gas and chondrules cool together as fast as they can either radiate away energy or leave the source of infrared radiation that heats them .
dc02 showed that the cooling rate of gas and chondrules should be @xmath18 where @xmath16 is the pre - shock gas density , @xmath19 the shock speed , @xmath17 is the concentration " of chondrules ( the chondrules - to - gas mass ratio normalized to @xmath20 , where a chondrule radius @xmath21 has been assumed ) , and @xmath22 the concentration of submicron dust ( the dust - to - gas mass ratio normalized to @xmath23 ) .
more precisely , @xmath22 should be interpreted as the rosseland mean opacity of the dusty gas , normalized to @xmath24 . for typical parameters , then , cooling rates @xmath25 are obtained , consistent with the cooling rates of chondrules .
the effect of line cooling the cooling of gas by emission of so - called line photons by trace molecules in the gas such as co and @xmath2 in solar nebula shocks has been considered by insn and by miura & nakamoto ( 2006 ) . in both cases
the cooling rates of nk93 were used .
insn assumed a gas optically thin to the line radiation and found chondrule cooling rates @xmath26 in many cases .
miura & nakamoto ( 2006 ) allowed the gas to become optically thick to this line radiation , using the formulation of nk93 described below .
they found that for plausible parameters the cooling rates of chondrules and gas cluster around @xmath27 , with line cooling playing an important role in the cooling of the gas .
chondrules and gas are thermally coupled only a few minutes after passing through the shock front ( insn ; dc02 ; ch02 ) , so these chondrule cooling rates are controlled by the rate at which gas cools by emission of line radiation
. to understand the essence of their results without reproducing all of their calculations in detail , we now estimate the cooling rate of chondrules and gas following the shock , using the tabulations of nk93 .
we first examine how the cooling rates predicted by nk93 and used by insn and miura & nakamoto ( 2006 ) are calculated . for a wide range of densities and temperatures ,
nk93 calculated the steady - state level populations of @xmath0 and co molecules ( denoted ` @xmath28 ' ) , and the rates of emission of photons .
the escape of these photons from a system was considered under the large - velocity - gradient ( sobolev ) approximation as a function of column density @xmath29 , via use of the parameter @xmath30 these results were adapted to a large range of geometries of interest , including those not consistent with the sobolev approximation , in particular the one of relevance to nebular shocks , a static , plane - parallel slab of thickness @xmath31 . according to nk93 , the cooling rate at the center of the slab ( at depth @xmath32 ) will depend on the similar parameter @xmath33 where @xmath34 is the thermal velocity of the molecules @xmath28 .
an important assumption made by nk93 is that if the line photons can escape to the edge of the slab , then the gas is cooled ; if the photons do not escape to the edge of the slab , they are essentially re - absorbed on the spot and there is no cooling .
the cooling rate per volume is then @xmath35 , where the subscript ` lte ' is the notation used by nk93 to denote cooling rates derived under the assumption that the population levels follow a boltzmann distribution , which is the regime of interest for solar nebula shocks .
nk93 have tabulated @xmath36 for both vibrational and rotational lines of @xmath0 and co , for every decade in @xmath37 .
the plane - parallel slab is equivalent to the nebula shock geometry because of the large velocity jump at the shock front . in a typical chondrule - forming shock ,
the gas velocity changes nearly instantaneously from @xmath38 to @xmath39 , a jump of over @xmath40 .
this velocity difference is much larger than the thermal velocities of the @xmath0 molecules [ @xmath41 @xmath42 at 2000
k ] that control the width of the line profile . essentially , once a line photon emitted from the post - shock gas escapes to the shock front , it will continue to travel relatively unimpeded by gas .
only one change to the nk93 formulation must be made before it is adopted to the nebular shock geometry . for gas a distance @xmath43 past the shock front
, the equivalent thickness of the slab is @xmath44 , and the cooling rate is only half the cooling rate calculated by nk93 , who assumed radiation could escape either side of the slab .
the cooling rate of the gas yields @xmath45 where we have ignored compression of the gas after passing through the shock front in this toy model , and we have assumed an abundance of he of 10% by number . converting the time derivative to a spatial derivative by assuming a constant gas velocity @xmath46 past the shock front , we find @xmath47 where we have assumed a constant @xmath34 throughout the post - shock region .
in actuality , @xmath34 would decrease past the shock front , but that would merely reduce the total cooling , so our assumption overestimates the degree of cooling .
this formula provides the gas temperature as a function of @xmath37 , the effective column density past the shock front .
fedkin & grossman ( 2006 ) showed that the @xmath48 ratio in the solar nebula probably varied with temperature , but found that @xmath49 was a good average for the temperatures of interest .
we have adopted a canonical ratio @xmath50 .
this is only slightly higher than some commonly assumed canonical ratios but is consistent with the abundance of lodders ( 2003 ) , @xmath51 .
we note that the exact value of @xmath48 in the chondrule formation environment is likely variable with time ( ciesla & cuzzi 2006 ) , and consider variations in this ratio later in the paper .
some dissociation of @xmath0 is likely to occur at the peak post - shock temperatures we consider , but we are interested in the maximum cooling possible by line photons , so we neglect dissociation in our calculations .
interpolating between the tabulated values of @xmath52 provided by nk93 , we have integrated equation [ eq : nkcool ] to find the total drop in temperature by the time the gas is at an equivalent of @xmath53 past the shock front . for typical parameters ( pre - shock density @xmath54 and velocity @xmath55 ; mixing ratio @xmath50 ; post - shock density @xmath56 and velocity @xmath57 ; and post - shock thermal velocity @xmath58 ) , this equates to @xmath59 , a distance @xmath60 , and a time 1.6 hours after passing through the shock front .
( again , this toy model neglects the compression and slowing of the gas .
it also assumes wrongly that @xmath34 is constant , whereas it should decrease slightly as the gas cools .
it is nonetheless illustrative . ) at these distances past the shock front , the total cooling due to rotational lines of @xmath0 and co is completely negligible , @xmath61 .
this illustrates that the rotational lines become optically thick so quickly that they are unable to escape the post - shock region and cool the gas . on the other hand ,
the total cooling from co vibrational line photons is roughly 72 k. this leads to an average cooling rate of @xmath62 , which is important , but not significantly greater than the cooling rates of chondrules ( @xmath63 ) .
the assumption of insn that the lines of @xmath0 and co are optically thin is therefore invalidated , because these lines are indeed optically thick .
the final line cooling mode ( not considered by insn , but considered by miura & nakamoto 2006 ) is vibrational line cooling from @xmath0 .
we find a total cooling due to these line photons on the order of 800 k. this is a significant cooling of the gas : the average cooling over 1.6 hours is @xmath64 , with even higher cooling rates obtained early on .
the cooling of the gas by vibrational line photons of @xmath0 , despite being somewhat optically thick , is the source of the high chondrule cooling rates predicted by miura & nakamoto ( 2006 ) .
we confirm the statement made by miura & nakamoto ( 2006 ) that `` ... line emission is important for the gas cooling . ''
we hereafter focus on cooling due to @xmath0 alone , as the dominant coolant is expected to be water . clearly , any model of chondrule cooling rates in solar nebula shocks must account for the emission of line photons from @xmath0 .
however , incorporating the cooling rates of nk93 directly into a shock code , as miura & nakamoto ( 2006 ) did , is not ideal .
first , line cooling due to rotational and vibrational transitions of @xmath0 was calculated by nk93 using roughly 50,000 transitions from the hitran database ( rothman et al .
while this was the best available data at the time , much more extensive databases now exist .
second , nk93 calculated escape probabilities under the large - velocity - gradient ( sobolev ) approximation , and asserted that the results will apply to the case of a static , plane - parallel slab if @xmath37 is defined as described above .
third , nk93 assumed a one - sided escape probability for line photons equal to @xmath65 , where @xmath66 is the sobolev optical depth ; more exact escape probabilities exist , as described below .
fourth , it is not necessarily the case that if photons do not escape to the edge of the slab that the gas does not cool locally ; photons re - absorbed halfway to the edge still cool the gas at the slab center . finally , nk93 did not consider the absorption of line photons by intervening dust .
inclusion of dust is absolutely mandatory , as the following argument makes clear : for a solar - composition gas with 0.5% of the mass in the form of @xmath67 radius grains ( similar to the size of matrix dust grains in meteorites ) , the opacity to short - wavelength radiation is as high as @xmath68 , where the internal density of the dust grains is @xmath69 . over the distance for which the optical depth in dust is unity ,
the parameter @xmath37 is therefore less than @xmath70 , assuming a ratio @xmath50 .
this corresponds to distances @xmath71 , or times of about 10 minutes , after passage through the shock front . as this is smaller than the @xmath37 for which the most significant cooling takes place , dust grains are capable of absorbing photons that would otherwise cool the gas .
a thorough treatment of line cooling should improve on the calculations of nk93 in the areas identified above . in this paper
we calculate the cooling rate of gas due to emission of line photons by @xmath0 molecules , improving upon the work of nk93 with the use of an expanded database , improved escape probabilities , and the inclusion of dust .
first we evaluate escape probability approximations , both in the presence and absence of dust .
we then calculate the cooling rates from @xmath0 molecules using the 1.2-million line scan-@xmath0 database ( jrgensen et al .
finally , we incorporate these cooling rates into a toy model of chondrule thermal evolution .
we conclude that for canonical parameters , line cooling is significant in the first few minutes past the shock front , but that afterward dust grains absorb the line photons that would otherwise cool the gas .
we discuss the effect of the important dust - to - water ratio on these conclusions .
we have used the scan-@xmath0 database of jrgensen et al . ( 2001 ) of 1.2 million lines ( rotational plus vibrational ) of the @xmath0 molecule to calculate the cooling rates due to emission of line photons from @xmath0 .
we continue to assume , as nk93 did , that if photons are not able to escape the region entirely , they are re - absorbed on the spot .
this assumption will be relaxed in future work , but for now it allows a crude estimate of the importance of line radiation . in other respects ,
we improve on the nk93 calculation by more accurately calculating escape probabilities , and by including the possibility of absorption by dust grains .
the probability that a line photon will escape a semi - infinite volume , otherwise known as the one - sided escape probability , @xmath72 , is given by @xmath73 @xmath74 @xmath75 ( avrett & hummer 1965 ; hummer & rybicki 1971 ; rybicki & lightman 1979 ; bowers & deeming 1984 ) . here
@xmath76(x ) is the line profile , @xmath77 is the frequency measured from line center in doppler widths ( @xmath78 ) , and @xmath79 is the second exponential integral . in equation [ eq : pesc ] , @xmath66 refers to the optical depth integrated over the line , which is @xmath80 times the optical depth at line center . considering the combined effect of both doppler and lorentz broadening , the line profile is given by the normalized voigt profile @xmath81 ( avrett & hummer 1965 ; rybicki & lightman 1979 ; bowers & deeming 1984 ) , where @xmath82 is the ratio of lorentz to doppler width ( avrett & hummer 1965 ) . in astrophysical situations
@xmath83 , perhaps as large as 0.1 ( mihalas 1978 ; bowers & deeming 1984 ) . in this study
, the voigt profile was calculated using the algorithm of zaghloul ( 2007 ) , where the voigt function is written as a single proper integral with a damped sine integrand .
a straightforward numerical integration was performed to calculate the
exact " one - sided escape probability @xmath72 in the absence of dust . besides the form adopted by nk93 , several other approximations to the one - sided escape probability exist in the literature .
hollenbach & mckee ( 1979 ) give the following approximation to the one - sided escape probability : @xmath84 the approximation given by collin - souffrin et al .
( 1981 ) is @xmath85 mathews ( 1992 ) estimates @xmath86 and dumont et al .
( 2003 ) give @xmath87 these approximations are all in the doppler limit , in which @xmath88 and the line profile is given by @xmath89 .
we have evaluated the approximations to the escape probabilities of hollenbach & mckee ( 1979 ) , collin - souffrin et al .
( 1981 ) , mathews ( 1992 ) , nk93 , and dumont et al .
( 2003 ) , and compared the results to our exact " escape probabilities ( figure 1 ) .
the maximum error between these approximations and the exact escape probability are as follows : 11.42% for hollenbach & mckee ( 1979 ) ; 15.70% for collin - souffrin et al .
( 1981 ) ; 31.34% for mathews ( 1992 ) ; and 37.22% for dumont et al .
the maximum error between the approximation used by nk93 and the exact escape probability is 46.44% , although it is not quite appropriate to make this comparison , as nk93 used the sobolov optical depth in their calculations .
the approximation of hollenbach & mckee ( 1979 ) is superior to the others and provides a satisfactory fit ; we will use this approximation in what follows . using the hollenbach & mckee ( 1979 ) approximation to the escape probability
, a parameter study was conducted to determine the effect of varying the value of @xmath82 in the calculation of the voigt function .
it was found that varying the value of @xmath82 between 0.0 and 0.1 increased the escape probability significantly for @xmath90 ( figure 2 ) , consistent with figures 4 and 5 of hummer & rybicki ( 1982 ) .
it is not immediately clear how this will affect the overall cooling rate , but we anticipate a small effect as the lines at which the escape probability increases with increasing @xmath82 are optically thick , leading to lower cooling rates .
we investigate this effect of the voigt function on the cooling rates in section 3 .
dust can also absorb photons , preventing their escape .
hence , the proper treatment of the escape probability must account for the absorption by dust , where the optical depth to line photons is now given by @xmath91 , and the exact escape probability is @xmath92 this equation reflects the fact that dust grains provide a continuum opacity that is always capable of absorbing photons .
we have used this equation to calculate the exact escape probabilities in the presence of dust ( figure [ fig : sprinkler ] ) . because it is computationally burdensome to calculate the escape probability for all possible combinations of @xmath66 and @xmath93 , we seek an approximation such that the escape probability is the product of two separate , independent functions of @xmath66 and @xmath93 . on physical grounds ,
one expects this function to resemble @xmath94 .
clearly this is the appropriate expression when @xmath95 .
it is also exact in the limit @xmath96 .
the deviation is expected to be strongest when @xmath97 , but happily the deviation is not great . applying the slight modification , @xmath98 provides an excellent fit to the exact probability calculated using equation [ eq : pescdust ] , as illustrated in figure [ fig : sprinkler ] , and tabulated in table 1 .
lr 0.0&11.42 + 0.03&24.52 + 0.1&25.02 + 0.3&47.51 + 1.0&54.27 + 3.0&47.73 + 10.0&65.20 + 30.0&128.04 + it can be seen that the approximations including dust fit very well in all regions except where @xmath99 and @xmath100 , as expected .
we can apply these escape probabilities to the calculation of the cooling rate from water when dust is present .
the cooling per line in the optically thin limit is given by @xmath101 where @xmath102 is the temperature - dependent line strength of the spectral line , and @xmath103 is the wavelength at line center ( see appendix ) . in our calculations , we assume that level populations obey a boltzmann distribution .
this is justified if the number density of protoplanetary disks greatly exceeds the critical density of each transition . as nk93 found no variations in level populations when the density exceeded @xmath104 ,
we assume that all level populations have critical densities below this value . since we are motivated by chondrule formation , for which the gas densities exceed @xmath105 , the use of a boltzmann distribution is justified .
unfortunately , for other problems ( e.g. , molecular clouds ) , our approach is limited in applicability .
the cooling rate is reduced below its optically thin limit as the column densities are increased , due to the inability of photons to escape the system .
specifically , the total cooling , @xmath106 , is given by the total cooling per line , summed over all lines , including the escape probabilities : @xmath107 based on the methods of plume et al .
( 2004 ) , the gas optical depth at line center , @xmath108 , can be found from the linestrength and column density of water : @xmath109 ( see appendix ) , where the doppler linewidth is given by @xmath110 and is @xmath111 at @xmath112 . in terms of the optical depth at line center , @xmath113 .
the dust optical depth at a depth @xmath43 into a semi - infinite volume is given by @xmath114 where @xmath115 is the opacity per gram of gas . assuming a dust - to - gas ratio @xmath116 , a particle radius @xmath117 , and an absorptivity @xmath118 for @xmath119 and @xmath120 for @xmath121
, we derive @xmath122 \ , { \rm cm}^{2 } \ , { \rm g}^{-1 } \label{eq : ourkappa}\ ] ] we have determined the rosseland mean dust opacity to be @xmath123 at @xmath112 .
our estimates are similar to those derived by henning & stognienko ( 1996 ) using a particle size distribution instead of our simplified monodispersion .
we first calculate the cooling rates in the absence of dust . using many more transitions ( and improved escape probabilities )
we find that , in the absence of dust , our cooling rates are enhanced by about 30% over those calculated by nk93 ( figure [ fig : nk93comp ] ) . for the cooling without dust
, we found a slight difference ( @xmath124 ) in cooling rate between the cases where the escape probabilities are calculated exactly , and when we use the approximation of hollenbach & mckee ( 1979 ) .
for consistency , we will quote cooling rates obtained using the hollenbach & mckee ( 1979 ) approximation .
the cooling rate for the @xmath112 case , in the absence of dust , is plotted as the rightmost curve in figure [ fig : coolingrate ] .
also plotted in figure [ fig : coolingrate ] are the cases where the dust - to - water ratio is varied from 300 times its canonical value to 1/300 times its canonical value . here
the canonical case refers to a dust - to - gas mass ratio of 0.5% , yielding the opacity discussed above ; the dust - to - water ratio effectively measures only the dust opacity per gram of water vapor .
the canonical amount of water assumed here is a ratio @xmath125 .
a higher dust - to - water ratio means dust grains are more likely to absorb line photons , and the cooling rate due to line emission is reduced below the optically thin limit at a smaller total column density of water .
surprisingly , even small amounts of dust ( 300 times smaller than the canonical limit , or a dust - to - gas ratio of @xmath126 ) will prevent the gas from cooling as freely as it would without dust at column densities @xmath127 . the maximum error in each case between the exact value of the cooling rate ( using the escape probability of equation 15 ) and that found using the approximation to the escape probability ( equation 16 ) is shown in table 2 ( for @xmath112 ) .
figure [ fig : coolrate1500 ] gives the same information for the case when @xmath128 .
the cooling rates ( assuming @xmath88 , the pure doppler broadening case ) for @xmath129 k , 1500 k , 1750 k , 2000 k , and 2250 k , both in the absence of dust and with the canonical abundance of dust , are tabulated in tables 3 through 12 .
lr 0.0&5.54 + 1.0&14.01 + 3.0&13.29 + 10.0&12.24 + 30.0&11.04 + 100.0&9.26 + 300.0&7.45 + 1/3&14.77 + 1/10&15.64 + 1/30&15.39 + 1/100&17.93 + 1/300&18.40 + absorption of line photons by dust grains drastically affects the cooling of the gas .
the dust opacity is a strong function of wavelength , so we investigate the effects of assuming a constant opacity . in the continuum opacity case
it would be reasonable to assume the flux of photons propagating through a gas is equivalent to the flux derived assuming the gas had a wavelength - independent opacity equal to the rosseland mean opacity .
we therefore recalculated the cooling rate when the dust opacity equaled the rosseland mean opacity cited above , at all wavelengths .
for the canonical dust - to - water ratio , the results are plotted in figure [ fig : ross ] . the optical depth of dust
is given by @xmath130 using the rosseland mean opacity and canonical water - to - gas ratios cited above , we would estimate @xmath99 when @xmath131 .
this is indeed the point where the cooling rate is significantly reduced due to absorption by dust grains .
when the opacity is replaced by the wavelength - dependent opacity @xmath115 , however , the cooling is not effectively reduced until higher column densities are reached .
this signifies that most of the cooling is effected by emission of line photons with wavelengths at which @xmath132 , i.e. , @xmath133 .
use of a wavelength - dependent opacity in conjunction with a calculation of the cooling at each @xmath2 wavelength is therefore necessary .
we use the wavelength - dependent opacity in all cases cited here .
finally , we investigate the effect of the line broadening parameter @xmath82 on the cooling rate . increasing @xmath82 above the @xmath134 pure doppler broadening case
has the effect of putting more emission in the optically thin wings of the line profile .
effectively , this should act like a reduction of the overall water column density .
we have calculated the cooling rates assuming line profiles with @xmath88 and @xmath135 , in the cases where no dust is present , and when it is present at the canonical value .
these cooling rates are plotted in figure [ fig : acool ] .
as expected , the cooling rate is slightly higher when @xmath135 , because the column density of water is effectively reduced ; however , the difference is slight . in the case with no dust , the discrepancy between the @xmath88 and @xmath135 cases is @xmath136 for a column density of water of @xmath137 , increasing to @xmath138 at a column density of @xmath139 . in the case with dust ,
the discrepancy is @xmath140 at @xmath141 .
these calculations are much more computationally burdensome than the @xmath134 pure doppler broadening case , because of the need to calculate the voigt function .
fortunately , it is seen that their effects on the cooling rates of gas are minimal , @xmath142 in the regime where cooling is significant , for @xmath82 up to 0.1 .
as @xmath82 is typically @xmath143 ( mihalas 1978 ; bowers & deeming 1984 ) , we will assume @xmath88 in the cases that follow .
lccr 10@xmath144&1.547 x 10@xmath145&1.548
x 10@xmath145&0.03 + 10@xmath146&1.543 x 10@xmath145&1.546
x 10@xmath145&0.16 + 10@xmath147&1.517 x 10@xmath145&1.526
x 10@xmath145&0.61 + 10@xmath148&1.405 x 10@xmath145&1.425
x 10@xmath145&1.44 + 10@xmath149&1.128 x 10@xmath145&1.160
x 10@xmath145&2.82 + 10@xmath150&6.259 x 10@xmath151&6.415
x 10@xmath151&2.95 + 10@xmath152&2.227 x 10@xmath151&2.293
x 10@xmath151&2.96 + 10@xmath153&6.611 x 10@xmath154&6.816
x 10@xmath154&3.10 + 10@xmath155&1.615 x 10@xmath154&1.668
x 10@xmath154&3.33 + 10@xmath156&3.318 x 10@xmath154&3.437
x 10@xmath154&3.58 + 10@xmath157&5.750 x 10@xmath158&5.978 x 10@xmath158&3.95 + 10@xmath6&8.587 x 10@xmath159&8.959 x 10@xmath159&4.34 + lccr 10@xmath144&1.547 x 10@xmath145&1.548
x 10@xmath145&0.03 + 10@xmath146&1.543 x 10@xmath145&1.546
x 10@xmath145&0.16 + 10@xmath147&1.516 x 10@xmath145&1.525
x 10@xmath145&0.59 + 10@xmath148&1.397 x 10@xmath145&1.415
x 10@xmath145&1.26 + 10@xmath149&1.083 x 10@xmath145&1.097
x 10@xmath145&1.62 + 10@xmath150&4.686 x 10@xmath151&4.439
x 10@xmath151&5.27 + 10@xmath152&3.204 x 10@xmath154&2.765
x 10@xmath154&13.71 + 10@xmath153&1.459 x 10@xmath159&1.226 x 10@xmath159&15.98 + 10@xmath155&1.101 x 10@xmath160&2.148 x 10@xmath160&95.04 + 10@xmath156&@xmath161&@xmath161&0.00 + 10@xmath157&@xmath161&@xmath161&0.00 + 10@xmath6&@xmath161&@xmath161&0.00 + lccr 10@xmath144&2.704 x 10@xmath145&2.705
x 10@xmath145&0.02 + 10@xmath146&2.700 x 10@xmath145&2.704
x 10@xmath145&0.12 + 10@xmath147&2.667 x 10@xmath145&2.679 x 10@xmath145&0.47 + 10@xmath148&2.519 x 10@xmath145&2.548
x 10@xmath145&1.15 + 10@xmath149&2.127 x 10@xmath145&2.179
x 10@xmath145&2.47 + 10@xmath150&1.347 x 10@xmath145&1.383
x 10@xmath145&2.81 + 10@xmath152&5.597 x 10@xmath151&5.755
x 10@xmath151&2.84 + 10@xmath153&1.850 x 10@xmath151&1.905
x 10@xmath151&3.00 + 10@xmath155&4.992 x 10@xmath154&5.153
x 10@xmath154&3.26 + 10@xmath156&1.094 x 10@xmath154&1.132
x 10@xmath154&3.45 + 10@xmath157&1.917 x 10@xmath162&1.991 x 10@xmath162&3.86 + 10@xmath6&2.715 x 10@xmath158&2.837 x 10@xmath158&4.48 + lccr 10@xmath144&2.704 x 10@xmath145&2.705
x 10@xmath145&0.02 + 10@xmath146&2.699 x 10@xmath145&2.702
x 10@xmath145&0.12 + 10@xmath147&2.664 x 10@xmath145&2.677
x 10@xmath145&0.46 + 10@xmath148&2.502 x 10@xmath145&2.527
x 10@xmath145&1.01 + 10@xmath149&2.031 x 10@xmath145&2.055 x 10@xmath145&1.36 + 10@xmath150&9.849 x 10@xmath151&9.430 x 10@xmath151&4.26 + 10@xmath152&7.223 x 10@xmath151&6.338
x 10@xmath151&12.26 + 10@xmath153&3.081 x 10@xmath159&2.612
x 10@xmath159&15.23 + 10@xmath155&2.210 x 10@xmath160&4.353 x 10@xmath160&97.01 + 10@xmath156&@xmath161&@xmath161&0.00 + 10@xmath157&@xmath161&@xmath161&0.00 + 10@xmath6&@xmath161&@xmath161&0.00 + lccr 10@xmath144&4.219 x 10@xmath145&4.220
x 10@xmath145&0.01 + 10@xmath146&4.214 x 10@xmath145&4.217
x 10@xmath145&0.08 + 10@xmath147&4.177 x 10@xmath145&4.193
x 10@xmath145&0.37 + 10@xmath148&4.002 x 10@xmath145&4.040
x 10@xmath145&0.94 + 10@xmath149&3.501 x 10@xmath145&3.575
x 10@xmath145&2.11 + 10@xmath150&2.446 x 10@xmath145&2.511
x 10@xmath145&2.68 + 10@xmath152&1.158 x 10@xmath145&1.189
x 10@xmath145&2.74 + 10@xmath153&4.219 x 10@xmath151&4.343
x 10@xmath151&2.93 + 10@xmath155&1.240 x 10@xmath151&1.279
x 10@xmath151&3.15 + 10@xmath156&2.863 x 10@xmath154&2.958
x 10@xmath154&3.35 + 10@xmath157&4.913 x 10@xmath162&5.101
x 10@xmath162&3.83 + 10@xmath6&6.215 x 10@xmath158&1.156 x 10@xmath158&5.75 + lccr 10@xmath144&4.219 x 10@xmath145&4.220
x 10@xmath145&0.01 + 10@xmath146&4.213 x 10@xmath145&4.217
x 10@xmath145&0.08 + 10@xmath147&4.173 x 10@xmath145&4.188 x 10@xmath145&0.35 + 10@xmath148&3.973 x 10@xmath145&4.006
x 10@xmath145&0.82 + 10@xmath149&3.329 x 10@xmath145&3.365
x 10@xmath145&1.15 + 10@xmath150&1.754 x 10@xmath145&1.696
x 10@xmath145&3.27 + 10@xmath152&1.383 x 10@xmath145&1.233
x 10@xmath145&10.84 + 10@xmath153&5.735 x 10@xmath159&4.899
x 10@xmath159&14.57 + 10@xmath155&3.973 x 10@xmath160&7.831 x 10@xmath160&97.13 + 10@xmath156&@xmath161&@xmath161&0.00 + 10@xmath157&@xmath161&@xmath161&0.00 + 10@xmath6&@xmath161&@xmath161&0.00 + lccr 10@xmath144&6.083 x 10@xmath145&6.084
x 10@xmath145&0.01 + 10@xmath146&6.078 x 10@xmath145&6.082
x 10@xmath145&0.02 + 10@xmath147&6.040 x 10@xmath145&6.057
x 10@xmath145&0.28 + 10@xmath148&5.847 x 10@xmath145&5.893 x 10@xmath145&0.78 + 10@xmath149&5.250 x 10@xmath145&5.344
x 10@xmath145&1.79 + 10@xmath150&3.943 x 10@xmath145&4.045
x 10@xmath145&2.58 + 10@xmath152&2.084 x 10@xmath145&2.139
x 10@xmath145&2.68 + 10@xmath153&8.297 x 10@xmath151&8.536
x 10@xmath151&2.89 + 10@xmath155&2.617 x 10@xmath151&2.698
x 10@xmath151&3.12 + 10@xmath156&6.260 x 10@xmath154&6.464
x 10@xmath154&3.26 + 10@xmath157&1.013 x 10@xmath154&1.724
x 10@xmath154&3.83 + 10@xmath6&1.060 x 10@xmath154&1.118 x 10@xmath154&5.54 + lccr 10@xmath144&6.083 x 10@xmath145&6.084
x 10@xmath145&0.01 + 10@xmath146&6.077 x 10@xmath145&6.081
x 10@xmath145&0.06 + 10@xmath147&6.034 x 10@xmath145&6.051 x 10@xmath145&0.28 + 10@xmath148&5.803 x 10@xmath145&5.842
x 10@xmath145&0.69 + 10@xmath149&4.977 x 10@xmath145&5.025
x 10@xmath145&0.98 + 10@xmath150&2.783 x 10@xmath145&2.716
x 10@xmath145&2.39 + 10@xmath152&2.351 x 10@xmath145&2.129
x 10@xmath145&9.45 + 10@xmath153&9.721 x 10@xmath159&8.359
x 10@xmath159&14.01 + 10@xmath155&6.497 x 10@xmath160&1.283 x 10@xmath163&97.59 + 10@xmath156&@xmath161&@xmath161&0.00 + 10@xmath157&@xmath161&@xmath161&0.00 + 10@xmath6&@xmath161&@xmath161&0.00 + lccr 10@xmath144&8.284 x 10@xmath145&8.285
x 10@xmath145&0.01 + 10@xmath146&8.279 x 10@xmath145&8.283
x 10@xmath145&0.04 + 10@xmath147&8.241 x 10@xmath145&8.260
x 10@xmath145&0.22 + 10@xmath148&8.040 x 10@xmath145&8.092
x 10@xmath145&0.65 + 10@xmath149&7.364 x 10@xmath145&7.476
x 10@xmath145&1.52 + 10@xmath150&5.835 x 10@xmath145&5.979
x 10@xmath145&2.47 + 10@xmath152&3.383 x 10@xmath145&3.471 x 10@xmath145&2.63 + 10@xmath153&1.457 x 10@xmath145&1.499
x 10@xmath145&2.85 + 10@xmath155&4.870 x 10@xmath151&5.022 x 10@xmath151&3.11 + 10@xmath156&1.185 x 10@xmath151&1.222
x 10@xmath151&3.20 + 10@xmath157&1.726 x 10@xmath154&1.793
x 10@xmath154&3.87 + 10@xmath6&1.440 x 10@xmath162&1.543 x 10@xmath162&7.14 + lccr 10@xmath144&8.284 x 10@xmath145&8.285
x 10@xmath145&0.01 + 10@xmath146&8.279 x 10@xmath145&8.281
x 10@xmath145&0.04 + 10@xmath147&8.233 x 10@xmath145&8.250
x 10@xmath145&0.22 + 10@xmath148&7.976 x 10@xmath145&8.022
x 10@xmath145&0.58 + 10@xmath149&6.963 x 10@xmath145&7.023
x 10@xmath145&0.86 + 10@xmath150&4.065 x 10@xmath145&3.998
x 10@xmath145&1.64 + 10@xmath152&3.658 x 10@xmath145&3.361 x 10@xmath145&8.11 + 10@xmath153&1.532 x 10@xmath158&1.324
x 10@xmath158&13.56 + 10@xmath155&9.895 x 10@xmath160&1.954 x 10@xmath163&97.46 + 10@xmath156&@xmath161&@xmath161&0.00 + 10@xmath157&@xmath161&@xmath161&0.00 + 10@xmath6&@xmath161&@xmath161&0.00 +
line cooling is effective at cooling gas , provided the line photons are not reabsorbed by @xmath0 molecules or dust grains before they can escape to the cool , pre - shock region . in a nebular shock
, it is assured that sufficiently far from the shock front , no line photons can escape and the gas and dust will become thermally coupled and cool slowly , if at all .
what is not clear is the degree to which line cooling is significant in the region immediately past the shock .
does line cooling lead to a significant drop in temperature ( @xmath164 k ) before dust grains begin to reabsorb line photons ?
are the cooling rates of chondrules dominated by line cooling and therefore high ( @xmath165 ) at the temperatures at which chondrules crystallize , as found by insn , or can line cooling be neglected , as dc02 and ch02 assume ? to answer these questions we have constructed a toy model to assess the maximum possible importance of line cooling and to determine whether more detailed calculations of chondrule formation in nebular shocks need to include line cooling .
our toy model builds on the calculation presented in section 1 .
we find it convenient to convert a time derivative @xmath166 to a spatial derivative @xmath167 , and then convert the spatial variable @xmath43 ( distance past the shock front ) into a column density of water past the shock front ( assuming the water density remains constant ) .
then the cooling with column density of water is given by @xmath168 the cooling rate is a function of the water column density but also the dust - to - water ratio ( via @xmath169 ) . given a dust - to - water ratio , this equation
can then be integrated to find @xmath170 as a function of @xmath171 .
note that we have not accounted for the fact that the line broadening should decrease as the temperature drops , making it slightly harder for the gas to cool ; our simplified analysis therefore overestimates the cooling somewhat .
we plot the results of this integration in figure [ fig : chondrule ] for both the dust - free case and the case where the dust is one - tenth that of the canonical value ( in order to show the effect of even a small amount of dust ) .
for ease of comparison to chondrules we have also assumed the canonical water - to - gas ratio ( and other canonical values ) to convert @xmath171 to a time .
specifically , we assumed a pre - shock density of @xmath172 , a shock velocity of @xmath173 , assumed the density increased and the velocity decreased by a factor of @xmath174 past the shock front , and used our canonical water - to - gas ratio of @xmath175 .
these parameters are consistent with our assumption of an initial post - shock temperature of 2200 k ( dc02 ; ch02 ) . in the absence of dust
, the gas would cool below 1400 k in roughly 300 seconds , leading to a cooling rate of @xmath176 .
this cooling is attributable solely to effective cooling by water ; by comparison to nk93 we infer these are mostly vibrational photons . as chondrules and gas
are expected to be thermally well coupled more than 100 seconds past the shock front ( dc02 ; ch02 ) , this can be interpreted as a likely cooling rate of chondrules as well . the cooling rate begins to taper off as large column densities are reached , but not until times @xmath177 , by which time the gas and chondrules have traveled roughly @xmath178 past the shock front , equivalent to a column density of water @xmath179 . including the absorption of line photons by dust grains completely changes the thermal history of the gas and chondrules . using an opacity of dust @xmath180 ( at short wavelengths ) and a post - shock gas density of @xmath181 , the optical depth of dust exceeds unity after only 55 km past the shock front , which is reached after traveling only 48 seconds . during this first minute or so past the shock front , the line photons emitted by @xmath2 molecules escape freely and effectively cool the gas ; but after the first minute or so , they are absorbed by dust grains instead .
instead of @xmath0 line photons freely escaping to the shock front and cooling the gas and dust system , the photons are reabsorbed `` on - the - spot '' and do not cool the gas .
this effect of the reduced cooling rate is clearly seen in figure [ fig : chondrule ] beyond about @xmath182 seconds .
the trapping of line photons by dust grains becomes so effective , no further cooling by line emission is possible after a few minutes . ignoring the effects of cooling other than line photons
, the temperature would stabilize at about 1600 k. for this particular choice of parameters , chondrules would _ not _ cool at rates @xmath165 through their crystallization temperatures ( 1400 - 1800 k ) as in insn and miura & nakamoto ( 2006 ) ; in fact , they would nt cool at all under the assumptions of our toy model .
we have investigated the effect of varying the dust - to - gas and water - to - gas ratios on the cooling rates of chondrules .
their behaviors are similar to those depicted in figure [ fig : chondrule ] . in table 4
we report only the average cooling rates of chondrules over their crystallization temperature range ( i.e. , from 1800 k to 1400 k ) .
the gas density and shock velocity are not varied .
the results of this parameter study show that even a small amount of dust very effectively shuts down the line cooling due to h@xmath183o .
cccc 0.0 & 0.1 & 1275 & @xmath184 + 0.0 & 1.0 & 1000 & @xmath185 + 0.0 & 10.0 & 875 & @xmath186 + 0.1 & 0.1 & 2000 & n / a + 0.1 & 1.0 & 1650 & + 0.1 & 10.0 & 1225 & @xmath187 + 1.0 & 0.1 & 2175 & n / a + 1.0 & 1.0 & 2000 & n / a + 1.0 & 10.0 & 1600 & + 10.0 & 0.1 & 2200 & n / a + 10.0 & 1.0 & 2175 & n / a + 10.0 & 10.0 & 2000 & n / a +
we have built upon the line cooling work of nk93 , by using a much expanded database of spectral lines , improved escape probabilities , and the inclusion of dust grains .
we have found that although the cooling rate due to @xmath0 line cooling is a complicated combination of gas and dust opacity , the cooling rate can be well approximated by @xmath188 .
this allows us to investigate the maximum cooling effect @xmath0 line cooling will have in chondrule - forming nebular shocks . because vibrational line photons from @xmath0 so effectively cool the gas , the high oxygen fugacities and @xmath0 densities inferred for the chondrule - forming environment ( fedkin & grossman 2007 ) would seem incompatible with cooling rates @xmath189 100 k / hr , but we find that dust significantly reduces the photon escape probabilities and the cooling rates .
dust grains absorb the line photons that would escape and cool the gas . without dust ,
@xmath0 line cooling reduces the temperature of the gas to 1400 - 1800@xmath190k in @xmath191 0.1 hr and cooling continues at this rate ( @xmath19210@xmath193 k / hr ) . with dust , however , the grains absorb line photons and inhibit cooling , leading to a cessation of cooling after @xmath5 0.1 hr ( @xmath5 8 min ) .
line cooling is a very effective and important cooling agent in the first few minutes and must be included in any comprehensive shock model .
after the first few minutes , however , the thermally well - coupled gas and chondrules cool together at a slow rate ( only as fast as they travel @xmath5 1 optical depth through the dust ) , as in dc02 and ch02 .
insn failed to note this effect ; although they treat the gas and dust as well coupled ( t@xmath194 = t@xmath195 ) , their system was effectively optically thin to line emission , allowing all line photons to escape the region and cool the gas .
as we are interested at this time only in the maximum cooling effects of line photons ( the worst case scenario " ) , our calculations of cooling rates made no assumptions about radiative transfer ; the chondrule temperatures followed the gas temperature .
thermal histories of chondrules including radiative transfer , line cooling , and other relevant effects will be reported on in future work ( morris et al . 2009b , in preparation ) .
since we have assumed that line photons either escape completely or are absorbed immediately , the true situation will clearly be bracketed between the two extremes shown in figure [ fig : chondrule ] . in these calculations
, we have assumed that the dust is thermodynamically coupled to the gas .
all temperatures ( gas , dust , and chondrules ) , in actuality , will be determined by dust / chondrule opacity and radiative transfer .
dc02 showed that the cooling rates of chondrules ( without the inclusion of line cooling ) were given by the formula that appears above as equation 1 . in the case considered in figure
[ fig : chondrule ] , the chondrule concentration may be considered to be low ( @xmath196 ) , and the dust opacity is characterized by @xmath197 .
this yields a total cooling rate of gas and dust through the crystallization range of about @xmath198 after line cooling shuts off .
higher chondrule concentrations ( @xmath199 ) appear to have been typical during chondrule formation ( cuzzi & alexander 2006 ; alexander et al .
these would increase the cooling rate to several @xmath200 .
line cooling is significant during the first few minutes past the shock , and more detailed calculations will be necessary to constrain the initial drop in temperature from the peak to below the liquidus . however , our results have shown that dust effectively shuts down line cooling within minutes and it then takes hours for chondrules to crystallize in a shock , consistent with dc02 and ch02 .
as mentioned previously , we did not calculate line cooling due to co , as it is a factor of @xmath5 20 less effective than @xmath0 .
co also cools by emitting line photons , but dust will absorb these line photons as readily as those emitted by @xmath0 .
figure [ fig : chondrule ] shows a drop in temperature of @xmath5 600 k. if co were included , the temperature would probably drop an extra 5% ( @xmath5 630 k ) , which is within the uncertainties of the @xmath0 abundances we have assumed .
this study has shown that the gas and chondrules will cool much too rapidly to match the experimental constraints on chondrule cooling rates if dust is not present .
the question of the dust abundance and its opacity therefore becomes paramount .
but , significantly , we have found that enhancing only the amount of @xmath0 causes higher cooling rates , whereas enhancing the amount of @xmath0 and dust together does not .
what affects the cooling rate the most , therefore , is not the amount of water , but the dust - to - water ratio .
potential complications , though , include the possibility that dust will evaporate immediately upon passage of the shock , and the possibility that dust vapor may recondense in the post - shock region ( e.g. , scott & krot 2005 ) .
in a related vein , we note an application of this work only tangentially related to chondrule formation .
ciesla et al .
( 2003 ) suggested that phyllosilicates found in the fine - grained accretionary rims of chondrules in cm meteorites may have resulted from rapid gas - phase reactions of water vapor with silicates .
previous studies indicated that phyllosilicate production would be kinetically inhibited in the solar nebula ( prinn & fegley 1987 ) ; however , ciesla et al .
( 2003 ) showed that nebular shocks could result in a local increase in the water vapor pressure , thereby increasing the rate of phyllosilicate formation , as well as increasing the temperature at which phyllosilicates become stable .
chondrule - forming shocks in icy regions of the solar nebula could therefore account for both the formation of chondrules and their fine - grained phyllosilicate rims .
a possible objection to this hypothesis is that the high abundance of water vapor in the shocked region would cool the gas too rapidly to allow the production of much phyllosilicates .
as our work here shows , however , rapid cooling by line photons will not be associated with shocked , water - rich nebula gas , as long as the dust - to - water ratio is close to the canonical value for the solar nebula .
an investigation of this hypothesis is also planned for future work .
this work was supported by nasa origins of solar systems grant nng06gi65 g .
we would like to thank frank timmes for assistance on certain computing aspects of the problem .
alexander , c. m. o. , grossman , j. n. , ebel , d. s. , & ciesla , f. j. 2006 , science , 320 , 1617 amelin , y. , krot , a. n. , hutcheon , i. d. , & ulyanov , a. a. 2002 , science , 297 , 1678 anders , e. , & grevesse , n. 1989 , , 53 , 197 avrett , e. h. , & hummer , d. g. 1965 , , 130 , 295 bowers , r. l. , & deeming , t. 1984 , _ astrophysics .
volume 1 - stars _
( boston : jones and bartlett publishers , inc ) cernicharo , j. , & crovisier , j. 2005 , space science reviews , 119 , 29 ciesla , f. j. , & hood , l. l. 2002 , icarus , 158 , 281 ciesla , f. j. , lauretta , d. s. , cohen , b. a. , & hood , l. l. 2003 , science , 299 , 549 ciesla , f. j. , & cuzzi , j. n. 2006 , icarus , 181 , 178 collin - souffrin , s. , delache , p. , frisch , h. , & dumont , s. 1981 , , 104 , 264 connolly , h. c. , jr . , &
desch , s. j. 2004 , chemie der erde / geochemistry , 64 , 95 connolly , h. c. , jr .
, desch , s. j. , ash , r. d. , & jones , r. h. 2006 , meteorites and the early solar system ii , 383 cuzzi , j. n. , & alexander , c. m. o. 2006 , , 441 , 483 desch , s. j. , & cuzzi , j. n. 2000 , icarus , 143 , 87 desch , s. j. , & connolly , h. c. , jr .
2002 , meteoritics and planetary science , 37 , 183 desch , s. j. , ciesla , f. j. , hood , l. l. , & nakamoto , t. 2005 , chondrites and the protoplanetary disk , 341 , 849 dumont , a .- m . , collin , s. , paletou , f. , coup , s. , godet , o. , & pelat , d. 2003 , , 407 , 13 fedkin , a. v. , & grossman , l. 2006 , meteorites and the early solar system ii , 279 fedkin , a. v. , & grossman , l. 2007 , lunar and planetary institute conference abstracts , 38 , 2014 grossman , j. n. 1988 , meteorites and the early solar system , 680 henning , t. , & stognienko , r. 1996 , , 311 , 291 hewins , r. h. 1997 , annual review of earth and planetary sciences , 25 , 61 hewins , r. h. , & connolly , h. c. , jr .
1996 , chondrules and the protoplanetary disk , 197 hewins , r. h. , connolly , h. c. , lofgren , g. e. , jr . , & libourel , g. 2005 , chondrites and the protoplanetary disk , 341 , 286 hollenbach , d. , & mckee , c. f. 1979 , , 41 , 555 hummer , d. g. , & rybicki , g. 1971 , , 9 , 237 hummer , d. g. , & rybicki , g. b. 1982 , , 254 , 767 iida , a. , nakamoto , t. , susa , h. , & nakagawa , y. 2001 , icarus , 153 , 430 jones , r. h. , lee , t. , connolly , h. c. , jr . , love , s. g. , & shang , h. 2000 ,
protostars and planets iv , 927 jrgensen , u. g. , jensen , p. , srensen , g. o. , & aringer , b. 2001 , , 372 , 249 kita , n. t. , huss , g. r. , tachibana , s. , amelin , y. , nyquist , l. e. , & hutcheon , i. d. 2005 , chondrites and the protoplanetary disk , 341 , 558 levy , e. h. 1988 , meteorites and the early solar system , 697 liffman , k. , & brown , m. 1995 , icarus , 116 , 275 liffman , k. , & brown , m. j. i. 1996 , chondrules and the protoplanetary disk , pages 285 - 302 , 285 lodders , k. 2003 , meteoritics & planetary science , vol .
38 , supplement , abstract no.5272 , 38 , 5272 lugmair , g. w. , & shukolyukov , a. 2001 , meteoritics and planetary science , 36 , 1017 mathews , w. g. 1992 , , 386 , 90 mihalas , d. 1978 , san francisco , w. h. freeman and co. , 1978 .
650 p. , miura , h. , & nakamoto , t. 2006 , , 651 , 1272 neufeld , d. a. , & kaufman , m. j. 1993 , , 418 , 263 pilipp , w. , hartquist , t. w. , morfill , g. e. , & levy , e. h. 1998 , , 331 , 121 plume , r. , et al .
2004 , , 605 , 247 prinn , r. g. , & fegley , b. 1987 , annual review of earth and planetary sciences , 15 , 171 rothman , l. s. , et al .
1987 , , 26 , 4058 rybicki , g. b. , & lightman , a.
p. 1979 ,
_ radiative processes in astrophysics _ ( new york : wiley - interscience ) sanders , i. s. 1996 , chondrules and the protoplanetary disk , 327 scott , e. r. d. , & krot , a. n. 2005 , , 623 , 571 shang , h. , shu , f. h. , & lee , t. 1996 , astronomical society of the pacific conference series , 122 , 191 shu , f. h. , shang , h. , & lee , t. 1996 , science , 271 , 1545 shu , f. h. , shang , h. , glassgold , a. e. , & lee , t. 1997 , science , 277 , 1475 shu , f. h. , shang , h. , gounelle , m. , glassgold , a. e. , & lee , t. 2001 , , 548 , 1029 sorby , h .
c. 1877 , nature , 15 , 495 sobolev , v. v. 1960 , cambridge : harvard university press , 1960 , tachibana , s. , huss , g. r. , miura , h. , & nakamoto , t. 2004 , lunar and planetary institute conference abstracts , 35 , 1549 tachibana , s. , & huss , g. r. 2005 , , 69 , 3075 urey , h. c. 1967 , icarus , 7 , 350 urey , h. c. , & craig , h. 1953 , , 4 , 36 wadhwa , m. , & russell , s. s. 2000 , protostars and planets iv , 995 yu , y. , & hewins , r. h. 1998 , , 62 , 159 zaghloul , m. r. 2007 , , 375 , 1043
the cooling per line is given by the expression what we are given in the scan- line list is linestrength .
the 1.2 million lines of spectral data included in the line list cover a range in wavelength from @xmath5 6700 @xmath206 to 25 @xmath207 m .
the data consist of the following : the wavenumber at line center , @xmath208 , in @xmath209 , the temperature - independent line strength , @xmath210 , in km / mol , and the excitation energy , @xmath211 , in @xmath209 . from the temperature - independent line strength , @xmath210 , given in the list ,
the temperature - dependent linestrength , @xmath102 , is calculated as follows : + @xmath212 + the vibrational partition function , @xmath213 , is given for temperatures ranging from 200 k to 8000 k , in increments of 200 k , in the documentation for the scan - h@xmath183o line list .
we have interpolated values for @xmath213 using a quadratic spline .
the rotational partition function , @xmath214 , is calculated with a subroutine provided in the documentation for the scan - h@xmath183o line list .
@xmath102 , as shown by eq .
[ eq7 ] , is in units of km / mol , and we wish to have @xmath102 in @xmath215 @xmath216 hz , so we must convert wavenumber to frequency which gives | we consider gas at densities appropriate to protoplanetary disks and calculate its ability to cool due to line radiation emitted by @xmath0 molecules within the gas .
our work follows that of neufeld & kaufman ( 1993 ; , 418 , 263 ) , expanding on their work in several key aspects , including use of a much expanded line database , an improved escape probability formulism , and the inclusion of dust grains , which can absorb line photons .
although the escape probabilities formally depend on a complicated combination of optical depth in the lines and in the dust grains , we show that the cooling rate including dust is well approximated by the dust - free cooling rate multiplied by a simple function of the dust optical depth .
we apply the resultant cooling rate of a dust - gas mixture to the case of a solar nebula shock pertinent to the formation of chondrules , millimeter - sized melt droplets found in meteorites .
our aim is to assess whether line cooling can be neglected in chondrule - forming shocks or if it must be included .
we find that for typical parameters , @xmath0 line cooling shuts off a few minutes past the shock front ; line photons that might otherwise escape the shocked region and cool the gas will be absorbed by dust grains . during the first minute or so
past the shock , however , line photons will cool the gas at rates @xmath1 , dropping the temperature of the gas ( and most likely the chondrules within the gas ) by several hundred k. inclusion of @xmath0 line cooling therefore must be included in models of chondrule formation by nebular shocks . |
photoacoustic imaging ( pai ) is a novel promising tool for visualizing light absorbing structures in an optically scattering medium , which carry valuable information for medical diagnostics .
it is based on the generation of acoustic waves by illuminating an object with pulses of non - ionizing electromagnetic radiation , and combines the high contrast of pure optical and the high resolution of ultrasonic imaging .
the method has demonstrated great promise for a variety of biomedical applications , such as imaging of animals @xcite , early cancer diagnostics @xcite , and imaging of vasculature @xcite .
when an object is illuminated with short pulses of non - ionizing electromagnetic radiation , it absorbs a fraction of energy and heats up .
this in turn induces acoustic ( pressure ) waves , that are recorded with acoustic detectors outside of the object .
other than in conventional ultrasound imaging , where the source of acoustic waves is an external transducer , in pai the source is the imaged object itself .
the frequency bandwidth of the recorded signals is therefore generally broad and depends on the size and the shape of illuminated structures . throughout this paper
we assume a planar recording geometry , where the acoustic signals are recorded with omnidirectional detectors arranged on planes ( or lines ) , see figure [ fg : setup0 ] .
the planar geometry is of particular interest since it can be realized most easily in practical applications .
the recorded acoustic signals are then used to reconstruct the initially generated acoustic pressure which represents optically absorbing structures of the investigated object . for the planar recording geometry ,
two types of theoretically exact reconstruction formulas have been reported : temporal back - projection @xcite and fourier domain formulas @xcite .
numerical implementations of those formulas often lead to fast and accurate image reconstruction algorithms . in temporal back - projection formulas
the signals measured at time @xmath0 are back projected over spheres of radius @xmath1 with the detector position in the centre ( @xmath2 denotes the speed of sound ) . in fourier domain formulas
this back - projection is performed by interpolation in the frequency domain .
reconstruction methods based on fourier domain formulas are attractive since they reconstruct an @xmath3 image in @xmath4 floating point operations by using of the fast fourier transform ( fft ) .
straightforward implementations of back - projection type formulas , on the other hand , require @xmath5 operation counts , see @xcite .
the standard fft algorithm assumes sampling on an equally spaced grid and therefore , in order to implement the fourier domain formulas , interpolation in the fourier space is required .
interpolation in the fourier domain is a critical issue , and creates artifacts in reconstructed images , see the examples in section [ sec : num ] .
one obtains significantly better results by increasing the sampling density in the fourier space .
this is achieved by either zero - padding @xcite or by symmetrizing the recorded signals around @xmath6 ( which is equivalent to using the fast cosine transform instead of the fft ) . in this paper
, we propose an efficient reconstruction algorithm that uses the nonuniform ( or unevenly spaced ) fft @xcite and further increases the quality of reconstruction .
the nonuniform fft has been applied to a variety of medical imaging problems , such as standard x - ray ct , magnetic resonance imaging , and diffraction tomography @xcite , and has also been used implicitly in gridding algorithms @xcite .
all those algorithms deal with the problem of recovering a two ( or higher ) dimensional object function from samples of its multidimensional fourier transform on a non - cartesian grid .
our approach is conceptually quite different to the above mentioned references : the special structure of our problem allows to perform several one - dimensional nonuniform ffts instead of a single higher dimensional one .
this leads to a reduced numerical cost , compared to the above algorithms .
the proposed algorithm is more closely related to a reconstruction algorithm for x - ray ct suggested in ( * ? ? ?
* section 5.2 ) , which also evaluates the fourier transform on irregular samples by means of the one - dimensional fft .
this article is organized as follows : in section [ sec : problem ] we present the mathematical basics of fourier reconstruction in pai . in section [ sec : nufft ]
we review the nonuniform fft which is then used to derive the nonuniform fft based reconstruction algorithm in section [ sec : fra ] . in section [ sec
: num ] we present numerical results of the proposed algorithm and compare it with existing fourier and back projection algorithms .
the paper concludes with a discussion of some issues related to sampling and resolution in the appendix .
let @xmath7 denote the space of smooth functions with bounded support in the half space @xmath8 , @xmath9 . consider the initial value problem @xmath10 with @xmath11 . here
@xmath12 denotes the laplacian in @xmath13 and @xmath14 is the derivative with respect to @xmath0 .
we write @xmath15 , @xmath16 , @xmath17 , and define the operator @xmath18 by @xmath19 photoacoustic imaging for planar recording geometry is concerned with reconstructing @xmath20 from incomplete and possibly erroneous knowledge of @xmath21 . of practical interest are the cases @xmath22 and @xmath23 , see @xcite .
the operator @xmath24 can be inverted analytically by means of the exact inversion formula @xmath25 where @xmath26 , and @xmath27 denotes the @xmath28-dimensional fourier transform , @xmath29 equation ( [ eq : inv ] ) has been derived in @xcite for three spatial dimensions .
it can be proven in any dimension by using the inversion formula for the spherical mean radon transform of @xcite .
a related formula using the fourier cosine transform instead of the fourier transform has been obtained in @xcite for @xmath30 .
the inversion formula ( [ eq : inv ] ) yields an exact reconstruction of @xmath31 , provided that @xmath32 is given for all @xmath33 . in practical applications only a _ partial _ ( or _ limited view _ ) data set is available @xcite . in this paper
we assume that data @xmath34 are given only for @xmath35 , see figure [ fg : setup0 ] , which are modeled by @xmath36 where @xmath37 is a smooth nonnegative cutoff function that vanishes outside @xmath38 . using data ( [ eq : data - part ] ) , the function @xmath31 can not be exactly reconstructed in a stable way ( see @xcite ) .
it is therefore common to apply the exact inverse of @xmath24 to the partial data @xmath39 and to consider the result as an approximation of the object to be reconstructed .
more precisely , the function @xmath40 defined by @xmath41 is considered an approximation of @xmath31 .
the function @xmath40 is called _ partial fourier reconstruction_. fourier reconstruction algorithms in pai name numerical implementations of ( [ eq : inv2 ] ) . in this paper
we apply the one - dimensional nonuniform fft to derive a fast and accurate algorithm for implementing ( [ eq : inv2 ] ) .
the discrete fourier transform of a vector @xmath42 with respect to the nodes @xmath43 ( with @xmath44 even ) is defined by @xmath45(\omega_k ) : = \sum_{n=0}^{n-1 } e^{-i \omega_k n 2\pi / n } g_n \ , , \quad k = -n/2 , \dots , n/2 - 1 \,.\ ] ] direct evaluation of the @xmath44 sums in ( [ eq : dft ] ) requires @xmath46 operations . using the classical fast fourier transform ( fft ) this effort
can be reduced to @xmath47 operations .
however , application of the classical fft is restricted to the case of evenly spaced nodes @xmath48 , @xmath49 .
the one - dimensional nonuniform fft ( see @xcite ) is an approximate but highly accurate method for evaluating ( [ eq : dft ] ) at arbitrary nodes @xmath50 , @xmath51 , in @xmath52 operations . to derive the nonuniform fft
we closely follow the presentation of @xcite , which is based on the following lemma : [ le : fou ] let @xmath53 and @xmath54 .
assume that @xmath55 is continuous in @xmath56 $ ] , vanishing outside @xmath56 $ ] , and positive in @xmath57 $ ] .
then @xmath58 here @xmath59 denotes the one - dimensional fourier transform of @xmath60 .
[ thm : nufft ] let @xmath61 , @xmath62 , @xmath60 , and @xmath63 be as in lemma [ le : fou ] .
then , for every @xmath64 and @xmath65 we have @xmath66 with @xmath67 taking @xmath68 $ ] in ( [ eq : fourmont ] ) , gives @xmath69 and therefore @xmath70 interchanging the order of summation in the right hand side of the above equation shows ( [ eq : nufft - a ] ) , ( [ eq : nufft - b ] ) and concludes the proof .
in the following we assume that @xmath71 is an even number .
then @xmath72 where @xmath73 for @xmath74 , is an oversampled discrete fourier transform with the oversampling factor @xmath61 .
moreover we assume that @xmath63 is concentrated around zero and decays rapidly away from zero .
the nonuniform fft uses the formulas ( [ eq : nufft - a ] ) , ( [ eq : nufft - c ] ) to evaluate @xmath75 $ ] at the nodes @xmath50 .
the basic steps of the algorithm are as follows : 1 .
append @xmath76 zeros to the vector @xmath77 and evaluate @xmath78 , @xmath79 , in ( [ eq : nufft - c ] ) with the fft algorithm .
2 . evaluate the sums in ( [ eq : nufft - a ] ) approximately by using only the terms with @xmath80 , where the _ interpolation length _
@xmath81 is a small positive parameter .
since @xmath63 is assumed to decay rapidly , the truncation error in step ( ii ) is small .
@xmath82 @xmath83 @xmath84 @xmath85 @xmath86 @xmath87 the nonuniform fourier transform is summarized in algorithm [ alg : nufft ] .
all evaluations of @xmath60 and @xmath63 are precomputed and stored .
moreover , the classical fft is applied to a vector of length @xmath71 .
therefore the numerical complexity of algorithm [ alg : nufft ] is @xmath88 .
typically @xmath89 , in which case the numerical effort of the nonuniform fft is essentially twice the effort of the one - dimensional classical fft applied to an input vector of the same length .
see ( * ? ? ?
* section 3 ) for an exact operation count , and a comparison between actual computation times of the classical and the nonuniform fft . in our implementation
we choose for @xmath60 the _ kaiser bessel window _ , @xmath90 here @xmath91 is the modified bessel function of order zero .
the one - dimensional fourier transform of @xmath92 is @xmath93 if @xmath94 and @xmath95 otherwise .
the kaiser bessel window is a good and often used candidate for @xmath60 , since @xmath96 becomes extremely small for @xmath97 .
for example , with the parameters @xmath98 , and @xmath99 , we have for @xmath100 , @xmath101 and characteristic function of the interval @xmath57 $ ] . _
right : _ fourier transforms @xmath102 and @xmath103 in db ( decibel ) . here
db denotes the logarithmic decay @xmath104 of some quantity @xmath105.,title="fig:",scaledwidth=45.0% ] and characteristic function of the interval @xmath57 $ ] .
_ right : _ fourier transforms @xmath102 and @xmath103 in db ( decibel ) . here
db denotes the logarithmic decay @xmath104 of some quantity @xmath105.,title="fig:",scaledwidth=45.0% ] [ rem : nufft ] take @xmath106 and let @xmath60 be the characteristic function of the interval @xmath107 $ ]
. then @xmath108 and ( [ eq : nufft - a ] ) , ( [ eq : nufft - b ] ) reduce to the @xmath109 series @xmath110 which is a discretized version of shannon s sampling formula @xcite @xmath111 applied to the fourier transform of a function @xmath112 that vanishes outside @xmath113 $ ]
. see figure [ fg : decay ] for a comparison of @xmath109 and @xmath114 , with @xmath115 and @xmath116 .
one realizes that @xmath114 decays much faster than @xmath109 and is therefore much better suited for truncated interpolation .
in fact , @xmath117 for @xmath118 , whereas @xmath119 only for @xmath120 . an error estimate for the nonuniform fft using the kaiser bessel window is given in @xcite .
the result is @xmath121 for example , taking @xmath122 , @xmath123 and @xmath115 , the above error is as small as @xmath124 .
in this section we apply the one - dimensional nonuniform fft to photoacoustic imaging . throughout the following
we restrict our attention to two dimensions , noting that the general case @xmath9 can be treated in an analogous manner .
assume that @xmath31 is a smooth function that vanishes outside @xmath125 , and set @xmath126 , where @xmath37 is as in ( [ eq : data - part ] ) .
fourier reconstruction names an implementation of ( [ eq : inv2 ] ) , that uses discrete data @xmath127 with @xmath128 and reconstructs an approximation @xmath129 with @xmath130 . here @xmath40 is defined by ( [ eq : inv2 ] ) , @xmath44 is an even number and @xmath131 . in the appendix
we show that the sampling in ( [ eq : fd ] ) , ( [ eq : gd ] ) is sufficiently fine , provided that @xmath132 , where @xmath133 is the essential bandwidth of @xmath31 .
discretizing ( [ eq : inv2 ] ) with the trapezoidal rule gives @xmath134 where @xmath135 one notices that the inner sums in ( [ eq : dft-2d ] ) , @xmath136 can be exactly evaluated with @xmath44 one - dimensional ffts , and the outer sums @xmath137 can be approximately evaluated with @xmath44 one - dimensional nonuniform ffts . denoting the resulting approximation by @xmath138 and setting @xmath139 we finally find @xmath140 with the inverse two - dimensional fft algorithm .
@xmath82 @xmath83 @xmath141 @xmath142 @xmath143 @xmath144 @xmath145 @xmath146 @xmath147 @xmath148 the nonuniform fft based reconstruction algorithm is summarized in algorithm [ alg : nff ] .
its numerical complexity can easily be estimated . evaluating ( [ eq : fft-1a ] ) requires @xmath149 operations ( @xmath44 one - dimensional ffts ) , evaluating ( [ eq : fft-2 ] ) requires @xmath149 operations ( @xmath44 non - uniform ffts ) , and ( [ eq : fft-1 ] ) is evaluated with the inverse two - dimensional fft in @xmath150 operations .
therefore the overall complexity of algorithm [ alg : nufft ] is @xmath150 . in the next section
we numerically compare algorithm [ alg : nff ] with standard fourier algorithms presented in the literature @xcite , which all differ in the way how the sums in ( [ eq : fft-2 ] ) are evaluated : 1 . *
direct fourier algorithm .
* equation ( [ eq : fft-2 ] ) can not be evaluated with the classical fft algorithm because the nodes @xmath151 are non - equispaced .
the most simple way to evaluate ( [ eq : fft-2 ] ) is with direct summation .
because there are @xmath152 such sums , direct fourier reconstruction requires @xmath153 operations .
consequently it does not lead to a fast algorithm .
however , since ( [ eq : fft-2 ] ) is evaluated exactly , it is optimally suited to evaluate the image quality of reconstructions with fast fourier algorithms .
* interpolation based algorithm . * a fast and simple alternative to direct fourier
reconstruction is as follows : choose an oversampling factor @xmath154 and exactly evaluate @xmath155 at the uniformly spaced nodes @xmath156 , @xmath157 , with the one - dimensional fft algorithm . in a next step , linear interpolation
is used to find approximate values @xmath158 , see @xcite .
evaluating @xmath159 with linear interpolation requires @xmath160 operation and therefore the overall numerical effort of linear interpolation based fourier reconstruction is @xmath150 .
+ algorithms using nearest neighbor interpolation instead of the linear one have the same numerical complexity and have also been applied to pai ( see , e.g. , @xcite ) .
higher order polynomial interpolation has been applied in @xcite for a cubic recording geometry .
3 . * truncated @xmath109 reconstruction . *
if the function @xmath60 in algorithm [ alg : nff ] is chosen as the characteristic function of the interval @xmath161 $ ] , @xmath162 , then the nonuniform fast fourier transform reduces to the truncated @xmath109 interpolation considered in @xcite .
however , due to the slow decay of @xmath163 , truncation will introduce a non - negligible error in the reconstructed image ( see remark [ rem : nufft ] ) .
the fourier algorithms are also be compared with a numerical implementation of the back - projection formula @xmath164 where @xmath165 denotes the distance between the detector location @xmath166 and the reconstruction point @xmath167 .
equation ( [ eq : ubp ] ) has been obtained in @xcite by applying the method of descent to the three - dimensional universal back - projection formula discovered by xu and wang @xcite . again ( [ eq : ubp ] ) gives an exact reconstruction only if it is applied to complete data @xmath168 , @xmath169 . in the numerical experiments the back - projection formula
is applied to the partial data @xmath170 , see ( [ eq : data - part ] ) , and implemented with @xmath153 operation counts as described in ( * ? ? ?
* section 3.3 ) .
in the following we numerically compare the proposed nonuniform fft based algorithm with standard fourier algorithm and the back projection algorithm based on ( [ eq : ubp ] ) .
the cutoff function @xmath37 is constructed by convolution of @xmath171 with the characteristic function of @xmath172 ^ 2 $ ] , where @xmath173 is a small parameter and @xmath174 is chosen in such a way that @xmath175 .
typically , @xmath173 is chosen as a `` small '' multiple of the sampling step size @xmath176 . in all numerical experiments
we take @xmath177 , and @xmath178 . the window width @xmath62 is chosen to be slightly smaller than @xmath179 , where @xmath61 is the oversampling factor that determines the accuracy of the fourier reconstruction algorithms . ) .
_ bottom line : _ reconstruction with oversampling ( @xmath122).,title="fig:",scaledwidth=32.5%,scaledwidth=30.0% ] ) .
_ bottom line : _ reconstruction with oversampling ( @xmath122).,title="fig:",scaledwidth=32.5%,scaledwidth=30.0% ] + ) .
_ bottom line : _ reconstruction with oversampling ( @xmath122).,title="fig:",scaledwidth=32.5%,scaledwidth=30.0% ] ) .
_ bottom line : _ reconstruction with oversampling ( @xmath122).,title="fig:",scaledwidth=32.5%,scaledwidth=30.0% ] + )
. _ bottom line : _ reconstruction with oversampling ( @xmath122).,title="fig:",scaledwidth=32.5%,scaledwidth=30.0% ] ) .
_ bottom line : _ reconstruction with oversampling ( @xmath122).,title="fig:",scaledwidth=32.5%,scaledwidth=30.0% ] ( left ) and nonuniform fft based fourier algorithm ( right ) .
here white corresponds to function value 1 , black to function value -0.4 .
_ bottom line : _ difference images between direct and truncated @xmath109 reconstruction ( left ) , and direct and nonuniform fft based reconstruction ( right ) . here white ( resp .
black ) corresponds to function value @xmath180 ( resp .
@xmath181 ) . , title="fig:",scaledwidth=32.5%,scaledwidth=30.0% ] ( left ) and nonuniform fft based fourier algorithm ( right ) . here
white corresponds to function value 1 , black to function value -0.4 .
_ bottom line : _ difference images between direct and truncated @xmath109 reconstruction ( left ) , and direct and nonuniform fft based reconstruction ( right ) .
here white ( resp .
black ) corresponds to function value @xmath180 ( resp .
@xmath181 ) . , title="fig:",scaledwidth=32.5%,scaledwidth=30.0% ] + ( left ) and nonuniform fft based fourier algorithm ( right ) . here
white corresponds to function value 1 , black to function value -0.4 .
_ bottom line : _ difference images between direct and truncated @xmath109 reconstruction ( left ) , and direct and nonuniform fft based reconstruction ( right ) .
here white ( resp .
black ) corresponds to function value @xmath180 ( resp .
@xmath181 ) . , title="fig:",scaledwidth=32.5%,scaledwidth=30.0% ] ( left ) and nonuniform fft based fourier algorithm ( right ) .
here white corresponds to function value 1 , black to function value -0.4 .
_ bottom line : _ difference images between direct and truncated @xmath109 reconstruction ( left ) , and direct and nonuniform fft based reconstruction ( right ) .
here white ( resp .
black ) corresponds to function value @xmath180 ( resp .
@xmath181 ) . , title="fig:",scaledwidth=32.5%,scaledwidth=30.0% ] + ( left ) and nonuniform fft based fourier algorithm ( right ) .
here white corresponds to function value 1 , black to function value -0.4 .
_ bottom line : _ difference images between direct and truncated @xmath109 reconstruction ( left ) , and direct and nonuniform fft based reconstruction ( right ) .
here white ( resp .
black ) corresponds to function value @xmath180 ( resp .
@xmath181 ) . , title="fig:",scaledwidth=32.5%,scaledwidth=30.0% ] ( left ) and nonuniform fft based fourier algorithm ( right ) .
here white corresponds to function value 1 , black to function value -0.4 .
_ bottom line : _ difference images between direct and truncated @xmath109 reconstruction ( left ) , and direct and nonuniform fft based reconstruction ( right ) .
here white ( resp .
black ) corresponds to function value @xmath180 ( resp .
@xmath181 ) . , title="fig:",scaledwidth=32.5%,scaledwidth=30.0% ] as first case example we use a circular shaped object @xmath182 centered at @xmath183 ( see top left image in figure [ fg : fcirc-1 ] ) .
for such a simple object reconstruction artifacts can be identified very clearly . moreover , the data @xmath184 can be evaluated analytically ( see ( * ? ? ? *
equation ( b.1 ) ) ) as @xmath185 .\ ] ] here @xmath186 , @xmath187 is the principal branch of the complex logarithm , and @xmath188 $ ] denotes the real part of complex number @xmath189 .
the reconstruction results are depicted in figures [ fg : fcirc-1 ] and [ fg : fcirc-2 ] .
table [ tb : performance ] and figure [ fg : performance ] compare run times with the relative @xmath190-error @xmath191 were @xmath192 denotes the discrete image obtained by direct fourier reconstruction .
run times were measured for matlab implementations on a personal computer with 2.4 ghz athlon processor . in order to demonstrate the stability of the fourier algorithms
, we also performed reconstructions from noisy data , where gaussian noise was added with a variance equal to @xmath193 of the maximal data value .
the reconstruction results are depicted in figure [ fg : fcirc-3 ] .
.run times and error of different reconstruction methods . [ cols="<,^,^,^",options="header " , ] , scaledwidth=85.0% ] , title="fig:",scaledwidth=32.5%,scaledwidth=30.0% ] , title="fig:",scaledwidth=32.5%,scaledwidth=30.0% ] . ) .
_ second line : _ linear interpolation based reconstruction ( @xmath122 ) .
_ third line : _ reconstruction with back projection algorithm .
_ bottom line : _ reconstruction with nonuniform fft algorithm ( @xmath122 ) .
the horizontal profiles on the right are taken at @xmath194.,title="fig:",scaledwidth=32.5%,scaledwidth=30.0% ] ) .
_ second line : _ linear interpolation based reconstruction ( @xmath122 ) .
_ third line : _ reconstruction with back projection algorithm .
_ bottom line : _ reconstruction with nonuniform fft algorithm ( @xmath122 ) .
the horizontal profiles on the right are taken at @xmath194.,title="fig:",scaledwidth=32.5%,scaledwidth=30.0% ] + ) .
_ second line : _ linear interpolation based reconstruction ( @xmath122 ) .
_ third line : _ reconstruction with back projection algorithm .
_ bottom line : _ reconstruction with nonuniform fft algorithm ( @xmath122 ) .
the horizontal profiles on the right are taken at @xmath194.,title="fig:",scaledwidth=32.5%,scaledwidth=30.0% ] ) .
_ second line : _ linear interpolation based reconstruction ( @xmath122 ) .
_ third line : _ reconstruction with back projection algorithm .
_ bottom line : _ reconstruction with nonuniform fft algorithm ( @xmath122 ) .
the horizontal profiles on the right are taken at @xmath194.,title="fig:",scaledwidth=32.5%,scaledwidth=30.0% ] + ) . _
second line : _ linear interpolation based reconstruction ( @xmath122 ) .
_ third line : _ reconstruction with back projection algorithm .
_ bottom line : _ reconstruction with nonuniform fft algorithm ( @xmath122 ) .
the horizontal profiles on the right are taken at @xmath194.,title="fig:",scaledwidth=32.5%,scaledwidth=30.0% ] ) .
_ second line : _ linear interpolation based reconstruction ( @xmath122 ) .
_ third line : _ reconstruction with back projection algorithm .
_ bottom line : _ reconstruction with nonuniform fft algorithm ( @xmath122 ) .
the horizontal profiles on the right are taken at @xmath194.,title="fig:",scaledwidth=32.5%,scaledwidth=30.0% ] + ) .
_ second line : _ linear interpolation based reconstruction ( @xmath122 ) .
_ third line : _ reconstruction with back projection algorithm .
_ bottom line : _ reconstruction with nonuniform fft algorithm ( @xmath122 ) .
the horizontal profiles on the right are taken at @xmath194.,title="fig:",scaledwidth=32.5%,scaledwidth=30.0% ] ) .
_ second line : _ linear interpolation based reconstruction ( @xmath122 ) .
_ third line : _ reconstruction with back projection algorithm .
_ bottom line : _ reconstruction with nonuniform fft algorithm ( @xmath122 ) .
the horizontal profiles on the right are taken at @xmath194.,title="fig:",scaledwidth=32.5%,scaledwidth=30.0% ] + + in the next example we consider the shepp
logan phantom @xmath195 , which is shown in top left image in figure [ fg : fphant ] .
the data were calculated numerically by implementing dalemberts formula @xcite , @xmath196 with @xmath197 denoting the spherical mean radon transform of @xmath198 .
the reconstruction results from simulated data are depicted in figure [ fg : fphant ] .
we emphasize that none of the above fourier algorithms are designed to calculate an approximation of @xmath31 but an approximation to the partial fourier reconstruction @xmath40 defined in ( [ eq : inv2 ] ) .
therefore even in the direct reconstruction ( top right image in figure [ fg : fcirc-2 ] ) and in the back projection reconstruction one can see some blurred boundaries in the reconstruction .
such artifacts are expected using limited view data ( [ eq : data - part ] ) , see @xcite .
the results of interpolation based reconstruction without oversampling ( @xmath106 ) are quite useless .
the reconstructions are significantly improved by using a larger oversampling factor @xmath61 .
however , even then , the results never reach the quality of the nonuniform fft based reconstruction .
moreover , the numerical effort of linear interpolation based reconstruction is proportional to the oversampling factor , which prohibits the use of `` very large '' values for @xmath61 ( see figure [ fg : performance ] ) . in the reconstruction with @xmath122 ( bottom line in figure [ fg : fcirc-1 ] and middle line in figure [ fg : fphant ] ) artifacts are still clearly visible .
the images in the middle line of figure [ fg : fcirc-2 ] suggest that truncated @xmath109 and nonuniform fft based reconstruction seem to perform quite similar .
however , the differences to the direct fourier reconstructions , shown in the bottom line in figure [ fg : fcirc-2 ] , demonstrate the higher accuracy of the nonuniform fft based algorithm .
the results in figure [ fg : fcirc-3 ] show that all applied reconstruction algorithms are quite stable with respect to data perturbation .
in particular , the filtered back projection algorithm produces images with the highest signal to noise ratio .
however , only at the cost of a nearly 100 times longer computation time ( see table [ tb : performance ] ) .
we presented a novel fast fourier reconstruction algorithm for photoacoustic imaging using a limited planar detector array .
the proposed algorithm is based on the nonuniform fft .
theoretical investigation as well as numerical simulations show that our algorithm produces better images than existing fourier algorithms with a similar numerical complexity .
moreover the proposed algorithm has been shown to be stable against data perturbations .
let @xmath31 be smooth function that vanishes outside @xmath38 , and define @xmath39 , @xmath40 by ( [ eq : data - part ] ) , ( [ eq : inv2 ] ) .
we further assume that @xmath199 is concentrated around zero and , that @xmath31 is essentially bandlimited with essential bandwidth @xmath133 , in the sense that @xmath200 is negligible for @xmath201 .
note that since @xmath31 has bounded support , @xmath202 can not vanish exactly on @xmath203 . *
* sampling of @xmath39 .
* equation ( [ eq : inv ] ) implies that @xmath204 with @xmath205 if @xmath206 and @xmath207 otherwise . the assumption that @xmath31 has essential bandwidth
@xmath133 and equation ( [ eq : shann - g ] ) imply that @xmath208 is negligible outside the set @xmath209 now shannon s sampling theorem @xcite states that @xmath39 is sufficiently fine sampled if the step size in @xmath210 and in @xmath0 satisfies the nyquist condition @xmath211 . *
* sampling of @xmath40 . * similar considerations as above
again show that @xmath40 is essentially bandlimited with essential bandwidth @xmath133 .
shannon s sampling theorem implies that @xmath40 can be reliable reconstructed from discrete samples taken with step size @xmath211 .
if @xmath31 has essential bandwidth larger than @xmath133 , the function @xmath39 has to be filtered with a low pass - filter _ before _ sampling .
otherwise , sampling introduces aliasing artifacts in the reconstructed image . theoretically , the resolution ( at least of the visible parts ) can be increased ad infinity by simply decreasing the sampling size @xmath212 . in practical applications , several other factors such as the bandwidth of the ultrasound detection system limit the bandwidth of the data , and therefore the resolution of reconstructed images @xcite
this , however , also guarantees that in practice a moderate sampling step size @xmath212 gives correct sampling without aliasing .
this work has been supported by the austrian science fund ( fwf ) within the framework of the nfn `` photoacoustic imaging in biology and medicine '' , project s10505-n20 .
moreover , the work of m. haltmeier has been supported by the technology transfer office of the university innsbruck ( transit ) .
x. d. wang , g. pang , y. j. ku , x. y. xie , g. stoica , and l. v. wang , `` noninvasive laser - induced photoacoustic tomography for structural and functional _ in vivo _ imaging of the brain , '' _ nature biotech .
_ , vol . 21 , no . 7 , pp .
803806 , 2003 .
r. a. kruger , k. d. miller , h. e. reynolds , w. l. kiser , d. r. reinecke , and g. a. kruger , `` breast cancer in vivo : contrast enhancement with thermoacoustic ct at 434 mhz - feasibility study , '' _ radiology _ , vol .
216 , no . 1 ,
279283 , 2000 .
s. manohar , a. kharine , j. c. g. van hespen , w. steenbergen , and t. g. van leeuwen , `` the twente photoacoustic mammoscope : system overview and performance , '' _ physics in medicine and biology _ , vol .
50 , no . 11 , pp . 25432557 , 2005 .
r. g. m. kolkman , e. hondebrink , w. steenbergen , and f. f. m. de mul , `` in vivo photoacoustic imaging of blood vessels using an extreme - narrow aperture sensor , '' _
ieee j. sel .
topics quantum electron .
_ , vol .
9 , no . 2 , pp . 343346 , 2003 .
r. o. esenaliev , i. v. larina , k. v. larin , d. j. deyo , m. motamedi , and d. s. prough , `` optoacoustic technique for noninvasive monitoring of blood oxygenation : a feasibility study , '' _ app .
_ , vol .
41 , no .
22 , pp . 47224731 , 2002 .
p. burgholzer , j. bauer - marschallinger , h. grn , m. haltmeier , and g. paltauf , `` temporal back - projection algorithms for photoacoustic tomography with integrating line detectors , '' _ inverse probl .
_ , vol . 23 , no . 6 , pp .
6580 , 2007 .
s. j. norton and m. linzer , `` ultrasonic reflectivity imaging in three dimensions : exact inverse scattering solutions for plane , cylindrical and spherical apertures , '' _ ieee trans . biomed .
_ , vol . 28 , no . 2 , pp . 202220 , 1981 . k. p. kstli , m. frenz , h. bebie , and h. p. weber , `` temporal backward projection of optoacoustic pressure transients using fourier transform methods , '' _ phys . med .
_ , vol . 46 ,
pp . 18631872 , 2001 .
d. potts , g. steidl , and m. tasche , `` fast fourier transforms for nonequispaced data : a tutorial , '' in _ modern sampling theory _ ,
anal.1em plus 0.5em minus 0.4emboston , ma : birkhuser boston , 2001 , pp .
247270 .
m. m. bronstein , a. m. bronstein , m. zibulevsky , and h. azhari , `` reconstruction in diffraction ultrasound tomography using nonuniform fft , '' _ ieee trans . med .
_ , vol . 21 , no . 11 , pp .
13951401 , 2002 .
g. paltauf , r. nuster , m. haltmeier , and p. burgholzer , `` experimental evaluation of reconstruction algorithms for limited view photoacoustic tomography with line detectors , '' _ inverse probl .
_ , vol . 23 , no . 6 , pp .
8194 , 2007 .
m. jaeger , s. schpbach , a. gertsch , m. kitz , and m. frenz , `` fourier reconstruction in optoacoustic imaging using truncated regularized inverse k - space interpolation , '' _ inverse probl .
_ , vol . 23 , pp .
s51s63 , 2007 . | fourier reconstruction algorithms significantly outperform conventional back - projection algorithms in terms of computation time . in photoacoustic imaging , these methods require interpolation in the fourier space domain , which creates artifacts in reconstructed images . we propose a novel reconstruction algorithm that applies the one - dimensional nonuniform fast fourier transform to photoacoustic imaging .
it is shown theoretically and numerically that our algorithm avoids artifacts while preserving the computational effectiveness of fourier reconstruction
. * key words . * image reconstruction , photoacoustic imaging , planar measurement geometry , fast algorithm , nonuniform fft . |
in 1996 september , a new x - ray transient was discovered with the wide field cameras ( wfcs ) on board the _ bepposax _ satellite and the source was designated sax j1808.43658 ( in t zand et al .
type i x - ray bursts were detected demonstrating that the compact object in this system is a neutron star and that the source is a low - mass x - ray binary ( lmxb ) . from the bursts , an initial distance estimate of 4 kpc was determined but this was later revised to 2.5 kpc ( in t zand et al .
1998 , 2001 ) .
the maximum luminosity of sax j1808.43658 was only approximately @xmath3 erg s@xmath4 indicating that the source was a faint neutron - star x - ray transient .
the outburst lasted about three weeks after which the source returned to quiescence . in april 1998 ,
the _ rossi x - ray timing explorer _ ( _ rxte _ ) satellite scanned , using the proportional count array ( pca ) , the region on the sky where sax j1808.43658 is located and the source was found to be active again ( marshall 1998 ) . follow up _
rxte_/pca observations showed that sax j1808.43658 exhibits coherent millisecond x - ray oscillations with a frequency of approximately 401 hz ( wijnands & van der klis 1998 ) , making this source the first ( and so far the only ) accretion - powered millisecond x - ray pulsar discovered .
the source was also detected in the optical ( giles , hill , & greenhill 1999 , with an optical position of r.a .
= 18@xmath5 08@xmath6 27.54@[email protected]@xmath9 and dec .
36 58@xmath1 44.3@[email protected]@xmath10 ; all coordinates in the paper are for j2000.0 ) and in the radio ( gaensler , stappers , & getts 1999 ) .
also this outburst lasted only for a few weeks ( e.g. , gilfanov et al .
1998 ) . during quiescence ,
sax j1808.43658 was observed with the narrow field instruments ( nfi ) on board _ bepposax _ and near the position of sax j1808.43658 a faint ( a few times @xmath11 erg s@xmath4 ) x - ray source was discovered ( stella et al .
2000 ) , which was identified with the pulsar based on presumed positional coincidence . an _ asca _
observation about half a year later resulted in a detection for sax j1808.43658 in quiescence but with a luminosity at least 4 times lower than measured with _
( dotani , asai , wijnands 2000 ) .
this low luminosity makes sax j1808.43658 the dimmest quiescent neutron - star lmxb x - ray transient detected so far ( see , e.g. , asai et al . 1996 , 1998 ) .
sax j1808.43658 stayed dormant until 2000 january 23 , when _ rxte _ found the source to be active again ( van der klis et al 2000 ) . due to solar constraints ,
the source was unobservable before this date ( from the end of november 1999 ) and it is unclear when exactly the source had become active .
the source was only detected at low luminosities ( up to a few times @xmath12 erg s@xmath4 ) with highly variable activity lasting up to 13 may 2000 ( wijnands et al . 2000 , 2001 ) .
here we report on _
bepposax_/nfi observations during this outburst .
to study the 0.2100 kev x - ray spectrum of sax j1808.43658 , we had _
bepposax_/nfi target - of - opportunity observations planned during its anticipated 2000 outburst .
these observations were triggered and we obtained 3 observations in march 2000 ( see tab . [
tab : log ] ; for a total of 89.1 ksec ) .
the data were analyzed using several different approaches , but here we report only on the maximum - likelihood ratio ( mlr ) method which is the most sensitive one ( the results of the other approaches were consistent with the results obtained from the mlr analysis ) .
the details of this method are described in kuiper et al .
( 1998 ) and in t zand et al .
( 2000 ; see also , e.g. , kuulkers et al . 2000 ) .
as already mentioned above in section [ section : introduction ] , the source was highly variable during the 2000 outburst ( see wijnands et al .
2001 ) and unfortunately all our _ bepposax_/nfi observations were taken during times when the source had very low luminosities and consequently only the data obtained with the mecs were useful . to optimize the sensitivity towards detecting sources
, we combined the data of the three observations .
the resulting mlr image is shown in figure [ fig : image ] . in the center of this image
four sources are detected ; their coordinates are listed in table [ tab : fov ] ( using the latest spatial calibration ; perri & capalbi 2001 , private communication ; the positions obtained by using the three observations individually varied only by @xmath010@xmath10 ) .
the brightest source in the field is the one closest to the position of sax j1808.43658 .
sufficient photons were detected from this source to extract its spectrum , which could satisfactorily be modeled with a power - law with photon index of [email protected] . assuming a column density @xmath13 of 1.22 @xmath14 @xmath15 ( using the @xmath16 from wang et al . 2001 and the relation between @xmath13 and @xmath17 from predehl & schmitt 1995 ) , the unabsorbed fluxes are [email protected] @xmath18 ( 0.55 kev ) and [email protected] @xmath18 ( 0.510 kev ) erg s@xmath4 @xmath15 . using a black - body
, we obtained a temperature @xmath19 of [email protected] kev and unabsorbed fluxes of [email protected] @xmath18 ( 0.55 kev ) and [email protected] @xmath18 ( 0.510 kev ) erg s@xmath4 @xmath15 .
however , the power - law fit is preferred over the black - body fit ( reduced @xmath20 is 1.1 vs. 1.9 , respectively ) . when fitting a two component model ( e.g. , power - law plus a black - body ) , the fit parameters could not usefully be constrained .
the position of this brightest source , however , is located 1.63@xmath1 from the position of the optical counterpart of the millisecond x - ray pulsar , which is considerably larger than the 0.55@xmath1 error radius ( 68% confidence level ( 1@xmath21 ) ; conservatively calculated as the direct sum of the statistical ( 18@xmath10 ) and systematical ( 15@xmath10 ) 1@xmath21 errors ) on the source position .
the millisecond x - ray pulsar is located at the 3@xmath21 location contour of this brightest _ bepposax _ source , and the probability that this source can still be identified with the millisecond x - ray pulsar is @xmath00.0029 .
although this would indicate that it is unlikely that this source can be identified with the millisecond pulsar , it is possible that the systematics in the position reconstruction for _ bepposax _ are not fully understood and we can not rule out completely the possibility that this source is indeed sax j1808.43658 . until it is conclusively demonstrated that this source is the pulsar or not
, we will refer to this source as `` the _ bepposax _ source '' .
we have reanalyzed the 1999 march 1719 _ bepposax_/nfi quiescent observation of sax j1808.43658 .
we confirm the presence of the source reported by stella et al .
( 2000 ) , however , the position of this source ( r.a .
= 18@xmath5 08@xmath6 32.9@xmath9 ; dec .
= 36 58@xmath1 07.5@xmath10 ; error radius 0.95@xmath1 , 68% confidence level ) and its count rates ( [email protected] @xmath22 counts s@xmath4 ; 1.310.0 kev ; see also stella et al . 2000 whom obtained very similar coordinates and count rates ) are almost identical to those obtained for the _ bepposax _ source during the combined 2000 march observations , suggesting that it is the same source ( the detected source is @xmath01.2@xmath1 away from the position of the pulsar ) . if this source can not be identified with the pulsar , then during all the _
bepposax_/nfi observations ( both the 1999 quiescent observation and the 2000 outburst ones ) , the millisecond x - ray pulsar was not detected .
we have reanalyzed the quiescent _ asca _ data of sax j1808.43658 ( dotani et al .
2000 ; see tab .
[ tab : log ] ) to investigate if the four detected _ bepposax _ sources are also detected with _
we used two different methods to analyze the _ asca _ data ; the method applied by dotani et al .
( 2000 ; hereafter referred to as the standard method ) , but also the mlr method .
the morphology of the _
asca_/sis images obtained with both methods ( see fig . [
fig : ascasis_image ] for the mlr image and dotani et al .
2000 for the image obtained with the standard method ) are very similar to each other , and a source was detected in the center of the images ( at @xmath23 significance , using the mlr method ) and its position is consistent with the position of sax j1808.43658 ( see tab .
[ tab : fov ] ; using an updated spatial calibration of _ asca _ ; gotthelf et al .
2000 ; we only list the positions obtained with the mlr method ) .
we also applied the mlr method to the combined gis2 and gis3 data .
although the coarse image binning and relatively high background prevented us from applying the standard analysis ( dotani et al .
2000 ) , the mlr method may be better suited to reconstruct such a noisy , coarse - rebinned image .
the resulting mlr image is shown in figure [ fig : ascasis_image ] .
we could detect the millisecond x - ray pulsar at a significance of @xmath24 .
the _ asca_/sis count rate ( 0.55.0 kev ) obtained using the mlr method is [email protected] @xmath22 counts s@xmath4 , which is significantly higher than the one we obtained using the standard method ( [email protected] @xmath22 counts s@xmath4 ) .
however , for faint sources it is important to correctly estimate the local background level at the source position , which is not trivial due to the extended _ asca _ point spread - function . to investigate the effect of a more structured background on the source count rate , we modeled the background near the pulsar in terms of including 3 extra point sources at `` hot '' positions in the mlr image , which are defined as the positions of remaining excesses in a residual mlr image after the detected sources were modeled out . using this background model ,
the count rate of the millisecond x - ray pulsar decreases to [email protected] @xmath22 counts s@xmath4 .
this count rate is still @xmath050% higher than obtained with the standard method , but the error bars are such that the count rates are consistent with each other . clearly , to obtain the correct _
asca_/sis count rates for sax j1808.43658 , it is very important to understand the local background level . using the mlr method
, we were able to extract a spectrum of the source .
however , the spectrum is of low statistical quality and multiple models could be fitted to the data equally well .
the spectrum can be fitted using a blackbody model ( resulting in @xmath19 of @xmath00.65 kev ) or with a power - law model ( with a photon index of @xmath25 ) .
we have also investigated the effects of the uncertainties in the column density @xmath15 . ] towards sax j1808.43658 by fitting the spectrum with an @xmath13 between 0.6 and 6 @xmath14 @xmath15 . a range of fluxes was obtained and we quote the typical unabsorbed fluxes , which include statistical errors ( as due to the fit procedure ) and systematic errors ( i.e. , the uncertainties in the background and in the column density ; using different spectral shapes like black - body or power - law ) : 7@xmath84 ( 0.55 kev ) and 10@xmath86 @xmath26 erg s@xmath4 @xmath15 ( 0.510 kev ) .
within the errors , those fluxes are consistent with what has been reported previously by dotani et al .
( 2000 ) . at the position of the _ bepposax _ source ,
no source could be detected , with a 3@xmath21 upper limit on the count rate of @xmath27 counts s@xmath4 ( 0.55.0 kev ; single sis ; depending on the spectral shape and the column density this would result in an unabsorbed 0.510 kev flux range of @xmath28 erg s@xmath4 @xmath15 ) .
two of the remaining three _ bepposax _ sources were detected during the _ asca _ observation ( fig .
[ fig : ascasis_image ] ) .
sax j1809.03659 was both detected in the sis ( although at the edge of the ccd detectors ; 5@xmath21 using the mlr method ) and gis detectors ( 4.5@xmath21 ) .
sax j1808.23654 was outside the field of view ( fov ) of the sis but inside that of the gis and was detected by this instrument ( 3.5@xmath21 ) .
the positions of those two sources are consistent with those obtained using _ bepposax _
[ tab : fov ] ) .
sax j1808.53703 was also outside the fov of the sis .
the gis did not detect the source .
an additional source was detected with the gis at a position of r.a .
= 18@xmath5 08@xmath6 02.8@xmath9 and dec .
= 37 02@xmath1 41.7@xmath10 ( @xmath29 ; designated ax j1808.03702 ) , although it might not be a point source but an extended source ( note that this causes additional systematical errors on the source position which can be as large as 2@xmath1 ) .
no pronounced counterpart of this source is visible in the _ bepposax_/mecs image ( fig .
[ fig : image ] ) , but it might be present at a 3@xmath21 level .
we have obtained _ bepposax_/nfi observations of the millisecond x - ray pulsar during its 2000 outburst . due to the extreme variability during this outburst ( see wijnands et al .
2001 ) , the _ bepposax _ observations were unfortunately taken when the source had very low luminosities .
we detected four field sources during our observations .
one source is 1.63@xmath1 away from the position of the millisecond x - ray pulsar , considerably larger than the 1@xmath21 error of 0.55@xmath1 .
this source can only be identified with sax j1808.43658 if we assume that the _ bepposax _ positional reconstruction is not completely understood .
if indeed this source can not be identified with the millisecond x - ray pulsar , then it should be designated sax j1808.63658 . a possible non - detection of sax j1808.43658 during our _ bepposax _ observations would result in an 3@xmath21 upper limit on the count rate of @xmath30 counts s@xmath4 ( 1.310 kev ; both mecs units active ) .
when assuming an @xmath13 of 1.22 @xmath31 @xmath15 and a power - law shaped spectrum with index 2 , this would result in an upper limit on the luminosity of @xmath32 erg s@xmath4 ( 0.510 kev ; for a distance of 2.5 kpc ) . in our _ asca _ observations we could not detect a source on the best position of the _ bepposax _ source with a range of upper limits on its x - ray flux of @xmath33 erg s@xmath4 @xmath15 ( 0.510.0 kev ; unabsorbed ) . if the column density is indeed @xmath34 cm@xmath4 then the upper limit determined with
( @xmath36 erg s@xmath4 @xmath15 ) was lower then flux detected during the _ bepposax _ observations ( [email protected] @xmath18 erg s@xmath4 @xmath15 ) .
this discrepancy can mean that indeed the _ bepposax _ source is the millisecond pulsar or also this source is variable . in the latter case , the fact that this source is detected during all _ bepposax_/nfi observations ( including the 1999 march 17 ) and not during the only _ asca _ observation should then be regarded as a coincidence .
the _ asca _
fluxes obtained ( using w3pimms ) for the other three _ bepposax _ sources were consistent with the values obtained from the _ bepposax _ fluxes ( taking into account the uncertainties in the spectra , the internal absorption of the sources , and in calculating the fluxes using w3pimms ) .
the nature of those field sources is unknown , but they could be , e.g. , agn . if indeed the _ bepposax _ source is a previously unknown source , its proximity to the millisecond pulsar raises the question if all the outbursts so far observed from the pulsar are indeed due to this source and not from this extra source .
it is clear that the source which exhibited the 1998 and 2000 outburst was the millisecond x - ray pulsar because during both outbursts the 401 hz pulsations were detected ( wijnands & van der klis 1998 ; van der klis et al .
furthermore , during both outbursts the optical counterpart increased considerable in luminosity ( giles et al .
1999 ; wachter & hoard 2000 ; wachter et al .
2000 ; note , that this secures the sub - arcsecond accurate position of the pulsar )
. however , the _ bepposax_/wfc positional errors of the persistent source detected during the 1996 outburst are consistent with both the position of the millisecond pulsar and the _ bepposax _ source . still , there are two measurements which provide convincing evidence that the 1996 outburst was also due to the millisecond x - ray pulsar .
first , the 99% error regions of the three bursts detected during that outburst are all consistent with the pulsar , while that of one burst is clearly inconsistent with the other source .
we note that all three bursts were observed in a small time window during the 1996 outburst and are most likely due to same source .
second , there is a marginal detection of a 401 hz oscillation during one of the three bursts ( in t zand et al .
2000 ) .
independent whether the _ bepposax _ source can be identified with the millisecond x - ray pulsar or that it is an unrelated field source , it is clear that sax j1808.43658 has very low luminosities in quiescence . as discussed in detail by wijnands
et al . ( 2001 ) , the very low luminosity of sax j1808.43658 during the _ bepposax _ observations demonstrate the highly variable nature of the source during its 2000 outburst .
for example , the source luminosity was below a few times @xmath11 erg s@xmath4 on 2000 march 56 and 8 , it reached @xmath37 erg s@xmath4 on 2000 march 11 as observed with the _ rxte_/pca , but the source was very dim again with _
bepposax_/nfi on 2000 march 2223 . also , on several occasions after 2000 march the source could again be detected with the _ rxte_/pca .
therefore , although the x - ray luminosity was very low during our _ bepposax _ observations , the clear long - term activity of the source demonstrated that these observations can not be regarded as true quiescent observations .
we reanalyzed the _
observation of sax j1808.43658 reported by dotani et al .
we confirm the detection of the millisecond x - ray pulsar . due to the uncertainties in its exact
_ asca _ count rate ( see section [ section : asca ] ) and its exact spectral shape , the exact quiescent luminosity of the millisecond pulsar is difficult to constrain but it is most likely in the range 0.31@xmath38 erg s@xmath4 ( 0.510 kev ) . despite this large uncertainty in its quiescent luminosity ,
the source is clearly fainter than the other neutron - star lmxb x - ray transients when they are in quiescence ( which are larger than a few times @xmath39 erg s@xmath4 ) .
if the quiescent luminosity of sax j1808.43658 is indeed as low as a few times @xmath40 erg s@xmath4 , then it is unclear whether the found distinction ( e.g. , garcia et al . 2001 and references therein ) between the quiescent luminosity of neutron - star transients and black - hole transients ( a few times @xmath41 erg s@xmath4 ) can hold . if certain neutron - star x - ray transients can indeed be as dim in quiescence as the black - hole systems , then the reasons to invoke the presence of event horizons in the black - hole systems will disappear ( garcia et al . 2001 ) .
more observations of neutron - star transients in quiescence are required to assess the validity of reported quiescent luminosity difference between the two types of systems .
such observations will also give important information to assess the question whether or not the millisecond x - ray pulsar is not only unique with respect to its pulsating nature , but also with respect to its quiescent behavior .
brown , bildsten , & rutledge ( 1998 ) proposed that the quiescent x - ray luminosities of neutron - star lmxb transients depend on their time - averaged mass accretion rates . compared with the other systems , the time - averaged accretion rate of the millisecond x - ray pulsar is low and brown et al .
( 1998 ) suggested that this source should have a low quiescent x - ray luminosity . within the uncertainties of their model and our observations ,
the predicted quiescent luminosity is consistent with the detected one .
it is also possible that this low quiescent luminosity is related to the fact that the source is so far the only accretion - driven millisecond x - ray pulsar known . for example , if this uniqueness is due to a higher neutron - star magnetic field than that of the other sources , then during quiescence the so - called propeller mechanism might inhibit residual accretion onto the neutron star . indications for residual accretion in quiescence have been found for the other neutron - star systems , which is mainly inferred from the power - law tail in their quiescent x - ray spectrum and the variability of several sources in quiescence ( however , see ushomirsky & rutledge 2001 for a discussion about x - ray variability due to deep crustal heating ) .
however , only at most @xmath050% of the quiescent x - ray flux of those sources is due to this power - law component , suggesting that the low quiescent luminosity of sax j1808.43658 can only in part be due to the lack of residual accretion in this system .
it is very well possible that the very low quiescent luminosity is due to a combination of the two above mentioned effects .
the low time - averaged accretion rate of sax j1808.43658 will not heat - up its neutron star to the same temperature as in the other systems , causing a lower thermal luminosity from its surface .
also , when residual accretion is inhibited due to its magnetic field , the power - law component is absent and the total quiescent x - ray luminosity of sax j1808.43658 could be very low compared to the other neutron - star x - ray transients in quiescence .
high quality quiescent x - ray spectra ( e.g. , with _
xmm - newton _ or _ chandra _ ) of sax j1808.43658 will help considerably to understand its low quiescent luminosity .
_ note added in manuscript : _
after submission of our paper , campana et al .
( 2001 ) reported preliminary results on their _ xmm - newton _ observation of sax j1808.43658 in quiescence .
they detected the source at a luminosity of approximately @xmath42 erg s@xmath4 ( 0.510 kev ) , consistent with our quiescent _
asca _ luminosity .
they also reported that at the position of the _ bepposax _ source no source could be detected .
however , the _ xmm - newton _ observation was not taken at the same time as our _ bepposax _ observations during the 2000 outburst and variability of this _ bepposax _ source can not be ruled out .
analysis of the ddt _ xmm - newton _ observation taken on 8 march 2000 , will definitely settled this issue because it was taken very close in time to our _ bepposax _ observations .
this work was supported by nasa through chandra postdoctoral fellowship grant number pf9 - 10010 awarded by cxc , which is operated by sao for nasa under contract nas8 - 39073 .
mk acknowledges support by the netherlands organization for scientific research ( nwo ) .
we thank jon miller and erik kuulkers for helpful discussions and carefully reading a previous version of this paper .
we thank the referee , josh grindlay , for useful comments on our paper .
campana , s. , stella , l. , gastaldell , f. , israel , g. , mereghetti , s. , burderi , l. , di salvo , t. 2001 , poster presented at `` new visions of the x - ray universe in the xmm - newton and chandra era '' ,
2630 november 2001 , noordwijk , the netherlands cccccc + sax j1808.43658 & optical & 18@xmath5 08@xmath6 27.54@xmath9 & 36 58@xmath1 44.3@xmath10 & 0.2@xmath10 & + & _ asca_/sis & 18@xmath5 08@xmath6 26.9@xmath9 & 36 58@xmath1 35.9@xmath10 & 0.24@xmath1 & [email protected] + & & & & & [email protected] + & _ asca_/gis & 18@xmath5 08@xmath6 29.8@xmath9 & 36 58@xmath1 27.6@xmath10 & 0.57@xmath1 & + _ bepposax _ source & _ bepposax _ & 18@xmath5 08@xmath6 35.1@xmath9 & 36 58@xmath1 07.8@xmath10 & 0.55@xmath1 & [email protected] + sax j1809.03659 & _ bepposax _ & 18@xmath5 09@xmath6 02.9@xmath9 & 36 59@xmath1 42.3@xmath10 & 0.56@xmath1 & [email protected] + & _ asca_/sis & 18@xmath5 09@xmath6 02.3@xmath9 & 36 59@xmath1 39.3@xmath10 & 0.23@xmath1 & [email protected] + & _ asca_/gis & 18@xmath5 09@xmath6 02.1@xmath9 & 36 59@xmath1 19.5@xmath10 & 0.53@xmath1 & + sax j1808.23654 & _ bepposax _ & 18@xmath5 08@xmath6 12.9@xmath9 & 36 54@xmath1 42.7@xmath10 & 0.59@xmath1 & [email protected] + & _ asca_/gis & 18@xmath5 08@xmath6 10.3@xmath9 & 36 54@xmath1 11.3@xmath10 & 0.63@xmath1 & [email protected] + sax j1808.53703 & _ bepposax _ & 18@xmath5 08@xmath6 32.0@xmath9 & 37 03@xmath1 03.1@xmath10 & 0.73@xmath1 & [email protected] + & _ asca_/gis & & & & @xmath431.6 + | we have observed the millisecond x - ray pulsar sax j1808.43658 on three occasions during its 2000 outburst with the _ bepposax _ satellite .
the source was highly variable and erratic during this outburst , and by coincidence we obtained data only during times when the source had very low luminosities . during our observations
, we detected four faint sources .
the source closest to the position of sax j1808.43658 is still @xmath01.6@xmath1 away .
this source can only be identified with sax j1808.43658 if we assume that the _ bepposax _ positional reconstruction is not completely understood .
we also reanalyzed a _
bepposax _ observation taken in march 1999 when the source was in quiescence and during which the source was thought to have been detected ( stella et al .
2000 ) . based on the similarities ( position and luminosity ) of this source with the above mentioned source @xmath2 away from sax j1808.43658 ,
it is possible that they are the same source .
if this source is not the millisecond pulsar , then during all _ bepposax _ observations of sax j1808.43658 ( the 2000 outburst ones and the 1999 quiescent one ) , the millisecond pulsar was not detected .
a reanalysis of the _ asca _ quiescent data of sax j1808.43658 ( dotani , asai , & wijnands 2000 ) confirms that during this observation the source was securely detected in quiescence .
we discuss our results for sax j1808.43658 in the context of the quiescent properties of low - mass x - ray binary transients . |
the diluted ising antiferromagnet fe@xmath8zn@xmath9f@xmath2 in an external magnetic field has proven to be a good model system for the random - exchange ising model ( reim ) ( @xmath10 0 ) and the random field ( rfim ) ising model ( @xmath11 0 ) @xcite . in the original derivation of the equivalence between the rfim and a diluted ising antiferromagnet in a uniform applied field
, weak dilution and small values of @xmath12 were assumed @xcite ( @xmath13 is the magnitude of the exchange interaction ) . in the fe@xmath8zn@xmath9f@xmath2 system ,
weak dilution implies fe concentrations well above the percolation threshold @xcite @xmath14=0.25 and the most convincing experimental results on the rfim critical behaviour have been obtained on samples with @xcite @xmath150.46 . on the other hand
, interesting dynamic properties may become observable in the limit of strong dilution .
rfim systems have been argued to attain extremely long relaxation times at temperatures near @xmath16 @xcite and for large values of @xmath12 , where the ordered phase is destroyed , it has been argued that a glassy phase will appear , even without exchange frustration being present in the system @xcite .
experimental results on fe@xmath8zn@xmath9f@xmath2 samples of concentration at or slightly above @xmath14 have revealed some dynamic properties similar to those of conventional spin glasses @xcite . in a recent paper @xcite
it was found that the percolating threshold sample fe@xmath17zn@xmath18f@xmath2 exhibits magnetic ageing , a typical spin glass feature , whereas the slowing down of the dynamics followed an arrhenius law , i.e. it did not support the existence of a finite temperature spin glass phase transition .
results using neutron scattering @xcite and faraday rotation technique @xcite have established random - field induced spin glass like dynamic behaviour in fe@xmath0zn@xmath1f@xmath2 .
recent magnetization experiments revealed that a similar behaviour occur at intense applied fields in samples of fe@xmath8zn@xmath9f@xmath2 , with @xmath19 = 0.56 and 0.60 @xcite . in this paper
we discuss experimental results from dc - magnetisation and ac - susceptibility measurements on the same fe@xmath0zn@xmath1f@xmath2 system of earlier neutron and faraday rotation measurements .
in zero applied field , a slowing down of the dynamics occurs at low temperatures that obeys a pure arrhenius law and some slowing down is also observable near the antiferromagnetic transition temperature . in applied dc - fields ,
additional slow dynamical processes are introduced near @xmath3 by the random fields .
a comprehensive static and dynamic phase diagram in the @xmath5 plane is deduced that , in parts , adequately compares with an earlier published phase diagram on the same compound @xcite .
a high quality single crystal @xcite of fe@xmath0zn@xmath1f@xmath2 in the form of a parallelepiped with its longest axis aligned with the crystalline @xmath20-axis was used as a sample .
the frequency dependence of the ac - susceptibility in zero applied dc - field was studied in a cryogenic ltd .
s600x squid - magnetometer .
a commercial lake shore 7225 ac - susceptometer was employed for the ac - susceptibility measurements in a superposed dc magnetic field and the temperature dependence of the magnetisation in different applied dc - fields was measured in a quantum design mpms5 squid - magnetometer .
the magnetic field was in all experiments applied parallel to the @xmath20-axis of the sample .
1 shows the temperature dependence of both components of the ac - susceptibility , ( a ) @xmath21(@xmath22 ) and ( b ) @xmath23(@xmath22 ) .
the different frequencies ranges from 0.051 - 51 hz as indicated in the figures .
the transition from a paramagnetic curie - weiss behaviour at high temperatures to long range antiferromagnetic order is signaled by the cusp in @xmath21(@xmath22 ) at about 20 k. a small bump in @xmath23(@xmath22 ) is observed at about the same temperature . below 15
k the ac - susceptibility becomes frequency dependent .
the out - of - phase component increases and a frequency dependent maximum that shifts towards lower temperatures with decreasing frequency is observed below @xmath24 5 k. the frequency dependence of @xmath21(@xmath22 ) and @xmath23(@xmath22 ) at low temperatures shows some resemblance with the behaviour of an ordinary spin glass . however , earlier neutron scattering measurements indicated that af lro is established below @xmath3 @xmath25 19.8 k in this system @xcite , provided the sample is submitted to a slow cooling process . to investigate the nature of the slowing down of the dynamics at low temperatures , a comparison is made with the behavior observed in ordinary spin glasses .
a 3d spin glass exhibits conventional critical slowing down of the dynamics @xcite according to : @xmath26 where @xmath27 is the microscopic spin flip time of the order @xmath28-@xmath29 s , @xmath30 the spin glass temperature and @xmath31 a dynamical critical exponent . defining the inflection point in @xmath23(@xmath22 ) as a measure of the freezing temperature @xmath32 for a relaxation time ( @xmath33 ) corresponding to the observation time , @xmath341/@xmath35 , of the ac - susceptibility measurement , the derived data may be employed for dynamic scaling analyses .
the data do not fit conventional critical slowing down according to eq .
1 with physically plausible values of the parameters . activated dynamics could govern the dynamics still yielding a finite phase transition temperature .
the slowing down of the relaxation times should then obey : @xmath36 where @xmath37 is a critical exponent @xcite .
the derived data fits eq .
2 with @xmath38 which implies that the slowing down rather is described by a generalized arrhenius law : @xmath39 fig . 2 shows the best fit to this expression yielding x=1 and @xmath27=10@xmath40 s for 0.051 @xmath41(hz ) @xmath42 1000 .
the observed frequency dependent ac - susceptibility shows striking similarities with the behaviour of alleged reentrant antiferromagnets .
in such a system there is a transition from a paramagnetic phase to an antiferromagnetic phase and spin glass behaviour is observed at low temperatures .
the reentrant ising antiferromagnet fe@xmath43mn@xmath44tio@xmath45 displays similar features as this system , e.g. the low temperature slowing down of the dynamics is found to obey a pure arrhenius behaviour @xcite . furthermore , the more diluted system fe@xmath17zn@xmath18f@xmath2 ( on the percolation threshold ) does not display long range antiferromagnetic order but it exhibits a slowing down of the relaxation times that follows a pure arrhenius law @xcite with a similar value of @xmath27 as here derived for fe@xmath0zn@xmath1f@xmath2 . in fig .
3 ( a ) @xmath21(@xmath46 ) and ( b ) @xmath23(@xmath47 ) are plotted for @xmath48=125 hz in different superposed dc magnetic fields @xmath49 2 t. at these rather low fields , the maximum in @xmath21(@xmath46 ) near @xmath3(@xmath50 ) gets rounded and is pushed towards lower temperature with increasing magnitude of the magnetic field .
the corresponding bump in the out - of - phase component in zero dc - field , increases in magnitude and sharpens for increasing dc - fields up to 1 t ( the inset of fig .
3 ( b ) displays fields up to 0.625 t ) .
a measure of the phase transition temperature @xmath3(@xmath50 ) is given by the position of the maximum in the derivative d(@xmath21@xmath51)/d@xmath51 @xcite . for fields @xmath7 1.5 t , @xmath3(@xmath50 ) is pushed to lower temperatures with increasing field strength following a reim to rfim crossover scaling , as described in ref .
13 . at higher fields
the maximum is washed out which signals that the antiferromagnetic phase is destroyed .
the destruction of the antiferromagnetic phase by strong random fields in fe@xmath8zn@xmath9f@xmath2 was observed by earlier faraday rotation @xcite and neutron scattering @xcite measurements in the same system ( @xmath19 = 0.31 ) , and by recent magnetization @xcite and dynamic susceptibility studies @xcite in less diluted samples ( @xmath19 = 0.42 , 0.56 and 0.60 ) .
a glassy dynamics is found in the upper portion of the @xmath5 phase diagram of fe@xmath8zn@xmath9f@xmath2 , at least within the interval @xmath52x@xmath53 . in increasing
applied dc - fields the out - of - phase component is enhanced in a rather narrow but widening region near the antiferromagnetic phase transition due to the introduction of random fields that create new slow dynamical processes in the system .
the increase of @xmath23(@xmath46 ) at lower temperatures , corresponding to the processes causing the slowing down of the dynamics already in zero field , remains observable also when the field is increased .
this latter feature can not be entirely attributed to random fields . for larger fields these low temperature processes and the processes caused by
the random fields start to overlap , and at the highest dc - fields they even become indistinguishable . in fig .
4 both components of the ac susceptibility are plotted , in an applied dc - field @xmath50=1.5 t , for @xmath48=15 , 125 and 1000 hz .
note that the temperature of the maximum in @xmath21(@xmath46 ) , at @xmath54 , shifts to lower temperatures as the frequency decreases . by way of contrast , no shift in the peak temperature is observable as a function of the frequency in dynamic susceptibility measurements performed in fe@xmath55zn@xmath56f@xmath2 @xcite and fe@xmath57zn@xmath58f@xmath2 @xcite , within the field limits of the weak rfim problem in each case .
the frequency dependent behaviour of @xmath54 is a feature associated with the effects of strong random fields in samples of fe@xmath8zn@xmath9f@xmath2 , particularly with @xmath19 close to @xmath59 . in fig . 5 ( a ) and ( b ) @xmath21(@xmath46 ) and @xmath23(@xmath60 ) are plotted for @xmath48=125 hz in different superposed dc magnetic fields @xmath612 t. the maximum in the in - phase - component is flattend , the susceptibility is strongly surpressed and the onset of the out - of - phase susceptibility is shifted towards lower temperatures as the dc - field is increased . no sign of a transition to an antiferromagnetic phase is observed .
6 shows the temperature dependence of the field cooled ( fc ) , @xmath62(@xmath51)/@xmath50 , and zero field cooled ( zfc ) , @xmath63(@xmath51)/@xmath50 , susceptibility @xcite at three different applied magnetic fields .
below a temperature @xmath64 the magnetisation becomes irreversible .
@xmath64 decreases with increasing field .
the irreversibility point is associated with an observation time mainly governed by the heating rate of the zfc experiment which in our experiment corresponds to about 100 s. in fig .
7 an @xmath5 magnetic phase diagram is shown , in which some of the above discussed experimental characteristics are summarised .
the open circles represent @xmath3(@xmath50 ) , the solid circles @xmath64(@xmath50 ) , diamonds the spin freezing temperatures @xmath32(@xmath50 ) for @xmath65=125 hz and open triangels label @xmath32(@xmath50=0 ) for different frequencies .
the onset of @xmath23(@xmath46 ) at frequencies @xmath48=15 , 125 and 1000 hz are shown as solid triangels , solid squares and open squares respectively .
those are measures that mirror the observation time dependence of @xmath64 . in diluted ising antiferromagnets
, @xmath3 is expected to decrease with increasing magnetic fields as : @xmath66 where @xmath67 is a crossover exponent and @xmath68 a small mean field correction . for low fields , @xmath491.5 t , we find @xmath691.4 using @xmath70=0 for @xmath3(@xmath50 ) as indicated by the solid line in fig .
7 . for higher fields , @xmath611.5 t
, a reversal of the curvature of @xmath64(@xmath50 ) occurs .
the dashed line corresponds to a functional behaviour according to eq .
4 with an exponenet @xmath69 3.4 .
a largely equivalent phase diagram has earlier been established for the same system utilising faraday rotation measurements @xcite .
one significant difference being that @xmath64(@xmath71 ) @xmath72(@xmath71 ) in ref .
@xcite , whereas we find a significant difference between these two temperatures , as is also observed in other dilute antiferromagnets @xcite .
the field dependence of @xmath3(@xmath50 ) is equivalent to those of the more concentrated fe@xmath55zn@xmath56f@xmath2 and fe@xmath73zn@xmath74f@xmath2 where the scaling behaviour of eq .
4 gives @xmath751.4 for fields up to 2 t and 10 t respectively @xcite .
the new features of the phase diagram in fig .
7 as compared to the one of ref .
@xcite are the observation time dependent spin freezing temperatures at low temperature and the observation time dependence of @xmath64(@xmath50 ) demonstrated by the shifts of the @xmath64(@xmath50 ) contours towards higher temperatures when decreasing the observation time . a possible mechanism for the spin freezing at low temperatures may be a weak frustration present in a third nearest neighbour interaction of this compound . results of numerical simulation @xcite indicates that a small frustrated bond plays no role in the reim properties of fe@xmath8zn@xmath9f@xmath2 under weak dilution .
however , it causes dramatic influences in the antiferromagnetic and spin glass order parameters close to the percolation threshold .
dynamic and static magnetic properties of the diluted antiferromagnet fe@xmath0zn@xmath1f@xmath2 have been studied .
the dynamic susceptibility in zero dc - field shows similarities to a reentrant ising antiferromagnet with a slowing down of the dynamics at low temperatures best described by a pure arrhenius law .
hence , there is no transition to a spin glass phase at low temperatures .
the field dependence of the antiferromagnetic transition temperature follows the predicted scaling behaviour for a random field system , in accord with earlier experimental findings@xcite .
the onset of @xmath76 occur above the antiferromagnetic phase transition , even in zero applied magnetic field .
@xmath76 shows a frequency dependent behaviour that mirror the observation time dependence of the fc - zfc irreversibility line .
the dynamics of the diluted antiferromagnet fe@xmath0zn@xmath1f@xmath2 has been shown to involve not only random field induced slow dynamics near @xmath3(@xmath50 ) , but additional slow dynamics originating from the strong dilution appears at low temperatures .
financial support from the swedish natural science research council ( nfr ) is acknowledged .
one of the authors ( fcm ) acknowledge the support from cnpq and finep ( brazilian agencies ) .
f. c. montenegro , k. a. lima , m. s. torikachvili , a. h. lacerda , j. magn .
magn . mater . *
177 - 181 * , 145 ( 1998 ) ; ibid . in _
magnetism magnetic materials and their applications _ edited by f. p. missel ( trans tech publications ltd , switzerland , 1999 ) , p. 371 . | the diluted ising antiferromagnet , fe@xmath0zn@xmath1f@xmath2 , has been investigated by dynamic susceptibility measurements in zero and and finite applied dc - fields . in zero field , there is a para- to antiferromagnetic phase transition at @xmath3 @xmath4 , followed by a dramatic slowing down of the dynamics at low temperatures .
the latter described by a pure arrhenius law .
the effect of a superposed dc - field on the antiferromagnetic phase transition and on the low - temperature dynamics is investigated , and a comprehensive static and dynamic phase diagram in the @xmath5 plane is derived . in agreement with earlier results on the same system , @xmath6follows a random - exchange ising model to random - field ising model crossover scaling for fields @xmath71.5 t. a random - field induced glassy dynamics appears for higher values of h , where the antiferromagnetic phase transition is destroyed .
the low - temperature dynamics shows striking similarities with the behavior observed in reentrant antiferromagnets .
0.2 cm pacs numbers : 75.40.gb 75.50.lk 0.2 cm 2 |
inflation well explains the origin of primordial density fluctuations , as well as flatness and homogeneity of our universe .
the general idea is so far quite successful , and inflationary models are confronted with precise observational data @xcite . since inflation is a high - energy phenomenon , it is important to study it in a more fundamental framework such as supergravity @xcite which is well motivated by particle physics and string theory . in particular , should tensor perturbations be detected in a near future , it would imply large ( trans - planckian ) excursions of the inflaton field @xcite . then the planck - suppressed corrections can not be neglected .
even if supersymmetry is broken at a higher scale than that of inflation , supergravity corrections have substantial impact on the scalar potential .
as is well known , a generic scalar potential in supergravity tends to be very steep in the large - field region , because of the exponential factor of the khler potential .
accordingly , it is hard to achieve flatness of the scalar potential along the whole inflationary trajectory in the case of large - field models .
therefore , some symmetries are usually imposed in the inflationary model building in supergravity .
a good example is the axion - like shift symmetry in the non - susy model @xcite and in the supergravity - based models @xcite .
the similar approach , assuming the global u(1 ) symmetry and the related monodromy structure of the superpotential , is known as the helical - phase inflation , because its inflaton is identified with the phase component of a complex scalar field rolling down a helicoid potential @xcite .
like in the more conventional shift - symmetric approach , a stabilizer superfield is used in the all known helical - phase inflationary models .
actually , the inflationary models with the non - compact u(1 ) and those based on the shift symmetry ( with or without a stabilizer superfield ) are equivalent , being related by the inflaton superfield redefinition @xmath7 .
still , it makes sense to study them separately because adding a simple symmetry - breaking or stabilizing term in one approach often leads to a complicated structure in the other approach . in the shift - symmetric approach
, a stabilizer superfield is needed to ensure positivity of the potential . in our previous work
@xcite ( see also ref .
@xcite ) , we proposed the alternative framework to achieve the same goal by stabilizing the scalar superpartner of the inflaton . in our approach ,
a number of the physical degrees of freedom was reduced , while a quartic shift - symmetric term was added to the khler potential . in this letter
, we study the helical - phase inflation without a stabilizer superfield . in sec .
[ sec : stabilize ] the radial part is stabilized by employing a higher - order ( polynomial ) term in the khler potential , similarly to refs .
the inflaton is identified with the pseudo - nambu - goldstone boson of the approximate u(1 ) symmetry of the khler potential .
a few particular models are studied in sec .
[ sec : models ] .
we conclude in sec .
[ sec : conclusion ] . throughout the paper ,
we use the natural ( reduced ) planck units , @xmath8 .
should the inflationary trajectory be along the phase direction in the helical phase inflation , the radial direction has to be constant during inflation .
for example , in the lln model it is achieved by taking the superpotential proportional to a negative power of the inflaton .
balancing the superpotential contribution diverging at the origin with the exponentially rising contribution due to the khler potential results in the stabilization of the radial part at a value of the order of the planck scale .
in our case without a stabilizer superfield , the inflaton potential includes both the superpotential itself and its derivative , and the formulae become rather complicated .
therefore , instead of dealing with a numerical minimization of the radial part , we employ the strong stabilization mechanism by using a higher order term in the khler potential , @xmath9 the first term is the usual ( minimal ) khler potential .
the constant term is added so that the expectation value of the khler potential approximately vanishes .
the second term is introduced for the purpose of stabilization .
thanks to that term , the radial part is stabilized at @xmath10 and minimizing it .
we assume the potential can be approximated as @xmath11 , and neglect derivatives of the superpotential since they are proportional to the slow - roll parameters .
as expected , the stabilized value of the radial part approaches to @xmath12 as @xmath13 goes to infinity .
moreover , @xmath13 of order one or smaller is sufficient for the truncation to be consistent in the case of @xmath14 .
more general khler potentials with similar features may exist , but we find the above example to be simple and efficient .
our stabilization mechanism is similar to those considered in the literature @xcite .
the stabilization parameter @xmath13 in eq . should be real and positive .
some comments about its magnitude are in order . when @xmath13 becomes large at a fixed @xmath15 , the khler metric ( the coefficient at the kinetic term ) may change its sign before reaching the symmetric phase , @xmath16 .
it occurs when @xmath17 for the khler potential in eq . .
then the above khler potential should be regarded as the effective description of the higgsed phase around @xmath18 .
it is enough for our purposes , since the radial part is stabilized throughout the process of inflation so we do not have to consider its dynamics .
when using the terms @xmath19 with @xmath20 , also allowed by the symmetry , in eq . , dynamics of the radial part can not be predicted once its distance from @xmath15 is more than that of the order one .
it may cause the typical problem of initial conditions for inflation .
we just assume here that the radial part is within the order - one distance from @xmath15 at the onset of inflation . ]
conversely , if @xmath13 becomes small at a fixed @xmath15 , the stabilized position of the radial part shifts inwards , @xmath21 , and eventually moves to the origin for @xmath22 .
depending on the value of @xmath13 , it is caused either by classical inflaton dynamics , quantum fluctuations , or quantum tunneling .
to avoid such situations , we take the value of @xmath13 to be of at least the same order as that of the critical value @xmath23 .
the stabilization strength can be measured by the mass of the radial part . when @xmath12 , the canonically normalized squared mass of the radial part is given by @xmath24 under the same conditions used for deriving eq . .
in the last inequality we also assume that @xmath25 and @xmath26 .
therefore , it is not difficult to strongly stabilize the radial part , _
i.e. _ @xmath27 . as long as the radial component is stabilized with its mass much larger than the hubble scale ,
the considerations in the next sections are independent upon the detailed mechanism of the stabilization .
after inflation , we can not neglect the derivatives of the superpotential , so that eqs . and are no longer valid .
the model - independent expression for the shift of the radial value from @xmath15 is very complicated , but it vanishes in the limit @xmath28 .
the radial component is not protected by any symmetry , contrary to the phase component .
we expect the mass of the radial component to be of at least the same order as the mass of the phase component . as will be clear in section [ sec : models ] , susy is generically broken spontaneously .
there is a susy breaking mass contribution @xmath29 near the vacuum , which becomes dominant in the large @xmath13 limit .
thus , the radial part is kept fixed near @xmath15 after inflation for a sufficiently large value of @xmath13 .
having stabilized the radial mode at @xmath30 , let us consider typical inflationary models , without introducing a stabilizer superfield .
let us parametrize the inflaton field as @xmath31 .
the phase is scaled so that it is canonically normalized .
the superpotential breaks the u(1 ) symmetry in the khler potential , and generates the inflaton ( scalar ) potential .
we study chaotic inflation with the quadratic potential , the starobinsky - like plateau potential , and a sinusoidal potential in this section .
the logarithmic singularity in the superpotential is the heart of the helical phase inflation , which is needed to realize a nontrivial spiral shape .
let us take the simplest ansatz @xmath32 where @xmath33 sets the scale of inflation , and @xmath34 ( with @xmath35 and @xmath36 real ) is the dimensional parameter controlling the cosmological constant .
after stabilization , the inflaton potential becomes @xmath37 with @xmath38 thus , the positive quadratic scalar potential is obtained under the condition @xmath39 .
the cosmological constant can be eliminated by choosing @xmath40 the full potential is shown in fig .
[ fig : quadratic ] for a limited field range . as is clear from the figure ,
the radial part is strongly stabilized , while its mass increases with the potential .
this is also implied by eq . . , @xmath41 , and @xmath42 .
the complex scalar is parametrized as @xmath43.,width=400 ] in this model , susy is not restored after inflation .
the gravitino mass at the vacuum is @xmath44 on the one hand , in the case of @xmath45 , the inflaton becomes much lighter than the gravitino , @xmath46 .
on the other hand , in the large @xmath15 limit , the inequality is reversed , @xmath47 .
the inflationary predictions of the model are well - known , and they do not change in our embedding .
nevertheless , a few comments are in order . by using the observed value of the amplitude of scalar perturbations , @xmath48 ,
the inflaton mass is determined as @xmath49 gev at the @xmath50-foldings @xmath51 .
the gravitino mass is roughly of the same order .
susy breaking is then supposed to be mediated to the mssm sector by the anomaly mediation , and is not compatible with the traditional low - energy susy scenario .
it is the typical consequence of removing a stabilizer field , see ref .
@xcite . in the last case ,
susy breaking at the intermediate scale @xcite can be motivated _
e.g. _ , by noticing that it stabilizes the electro - weak vacuum .
though the quadratic potential is already excluded by planck observations , some modifications or coupling to other sectors may make the quadratic model consistent with the data ( see _ e.g. , _ ref .
@xcite ) . instead of studying such possibilities
, we directly construct some viable inflationary models in the next subsections . in the previous subsection
, the logarithm @xmath52 in the superpotential leads to the quadratic potential .
a plateau - type potential consists of the exponential factors like @xmath53 , so let us consider the exponential of the logarithm , @xmath54 which is equivalent to the imaginary power of the superfield . in other words , let us take the following superpotential : @xmath55 where @xmath33 and @xmath56 are the constant parameters that determine the scale of inflation and the cosmological constant ( see below ) .
after stabilization , the inflaton potential becomes @xmath57 with the coefficients @xmath58 , \\
c= & \phi_0 ^ 2 -3 + \frac{1}{\phi_0 ^ 2}~~,\end{aligned}\ ] ] where the phase @xmath59 is defined by @xmath60 . for any @xmath15 larger than @xmath61 , @xmath62 and @xmath63 are positive definite , and the sign of @xmath64 depends on @xmath59 .
there exists a solution of @xmath59 such that @xmath64 is negative and , moreover , the cosmological constant vanishes .
though it is fully straightforward , the solution itself is not very illuminating and , hence , is not shown here .
the potential is a generalization of the starobinsky potential .
such potentials are often called `` starobinsky - like '' in the literature . our starobinsky - like scalar potential
is shown in fig .
[ fig : starobinsky ] . , @xmath41 , @xmath65 , and @xmath42 .
the complex scalar is parametrized as @xmath43.,width=400 ] the masses of inflaton and gravitino are given by @xmath66 the inflaton mass reads @xmath67 gev at @xmath68 . in this model ,
the inflaton is always lighter than the gravitino . in the limit @xmath69 , the inflaton mass approaches half of the gravitino mass .
the spectral index is the same as that of the starobinsky model , but the tensor - to - scalar ratio is different : @xmath70 in the leading order of @xmath71 . with @xmath26 ,
the tensor - to - scalar ratio is enhanced , when being compared to the starobinsky model ( @xmath72 ) . with an arbitrary imaginary power @xmath73 instead of @xmath74 in eq . , where @xmath75 is a real parameter , the tensor - to - scalar ratio is divided by @xmath76 as @xmath77 .
the previous examples are based on the superpotentials having the singularity at the origin . however , it is not the necessary feature of our mechanism because of the super - planckian value of the radial component .
let us take the superpotential of the previous subsection and replace its imaginary power by a real power as @xmath78 this is simply a linear function without the monodromy structure . in this case ,
a large value of @xmath79 is required not only by the positivity of the stabilized potential but also by the observational status of natural inflation .
after stabilization , the inflaton potential becomes @xmath80~,\end{aligned}\ ] ] with the coefficients @xmath81 and @xmath59 is again the argument of @xmath56 , @xmath82 .
the cosmological constant vanishes when @xmath83 ( 2 ) for the upper ( lower ) sign , and the sinusoidal scalar potential of natural inflation is obtained . the potential is shown in fig .
[ fig : natural ] .
, @xmath84 , @xmath42 , and @xmath85 , and the upper sign is taken in eq . .
the complex scalar is parametrized as @xmath43.,width=400 ] the masses of inflaton and gravitino are given by @xmath86 again , if the absolute value of the field is barely larger than the critical value @xmath87 ( this is for the lower sign ) , the inflaton is much lighter than the gravitino . in the large vev case
, gravitino becomes much lighter than the inflaton .
the parameter of the natural inflation is tightly constrained by the cmb observations .
the decay constant ( in our case @xmath88 ) must be larger than @xmath89 at 95% confidence level @xcite , so that the lower bound on the absolute value is obtained as @xmath90 .
when choosing @xmath91 , the inflaton mass is @xmath92 gev at @xmath68 .
in this paper we studied helical phase inflation with a single chiral superfield in supergravity , _
i.e. _ without the stabilizer superfield used in the known versions of helical phase inflation in the literature .
in order to ensure positivity of the scalar potential and avoid computational complexity , we introduced a stabilization term to the khler potential that fixes the radial component of the inflaton complex scalar at a sufficiently large value .
it results in technical simplification also . after the stabilization
, a slow - roll inflation occurs in the direction of the phase component .
we exemplified our findings on the three simple models of the single - superfield helical - phase inflation .
it implies that there should be many more possibilities to obtain viable inflationary potentials in our approach .
one such noticeable generalization is a hybrid version of the models in subsections [ ssec : plateau ] and [ ssec : natural ] .
let us take an arbitrary complex power of the inflaton superfield , @xmath93 with @xmath94 and @xmath75 real .
this model interpolates between the natural inflation and the starobinsky - like inflation .
a similar model was studied in the presence of the stabilizer superfield in ref .
@xcite .
although the radial part is stabilized at an over - planckian value , higher order terms in the khler potential are not necessarily problematic as long as we expand it around @xmath95 as in eq .
( see also footnote [ fn : awayfromminimum ] ) .
however , possible shift symmetry breaking terms in the khler potential may affect the inflaton potential and inflationary observables @xcite .
it is an interesting topic to be studied also in our setup elsewhere . in the models in subsections [ ssec :
quadratic ] and [ ssec : plateau ] , singularities and the monodromy structure are introduced in the superpotential , as in the lln approach .
this is beyond the usual field theory framework , and it is regarded as an effective description .
a possible uv completion of helical phase inflation in string theory was argued in ref .
@xcite . in summary , we proposed the new type of inflationary mechanism in supergravity , combining the ideas of helical phase inflation @xcite and single - superfield inflation with the higher dimensional stabilization term in the khler potential @xcite .
our models are simple : the kinetic term is approximately canonical , the superpotential is very economical , and no stabilizer superfield ( or extra d.o.f . )
is present .
the authors are grateful to the referee for his / her careful reading of the manuscript and critical remarks .
svk is supported by a grant - in - aid of the japanese society for promotion of science ( jsps ) under no .
26400252 , the world premier international research center initiative , mext , japan , the tmu special research fund and the competitiveness enhancement program of the tomsk polytechnic university in russia .
svk thanks the cern theory group for kind hospitality extended to him during preparation of this paper .
tt acknowledges the jsps for a grant - in - aid for jsps fellows and the grant - in - aid for scientific research no .
26@xmath9610619 .
99 p. a. r. ade _ et al .
_ [ bicep2 and planck collaborations ] , phys .
* 114 * ( 2015 ) 10 , 101301 [ arxiv:1502.00612 [ astro-ph.co ] ] .
p. a. r. ade _ et al .
_ [ planck collaboration ] , arxiv:1502.02114 [ astro-ph.co ] . d. z. freedman , p. van nieuwenhuizen and s. ferrara , phys . rev .
d * 13 * ( 1976 ) 3214 ; s. deser and b. zumino , phys .
b * 62 * ( 1976 ) 335 .
d. h. lyth , phys .
* 78 * ( 1997 ) 1861 [ hep - ph/9606387 ] .
k. freese , j. a. frieman and a. v. olinto , phys . rev .
* 65 * ( 1990 ) 3233 ; f. c. adams , j. r. bond , k. freese , j. a. frieman and a. v. olinto , phys .
d * 47 * ( 1993 ) 426 [ hep - ph/9207245 ] .
t. li , z. li and d. v. nanopoulos , phys .
d * 91 * ( 2015 ) 6 , 061303 [ arxiv:1409.3267 [ hep - th ] ] .
t. li , z. li and d. v. nanopoulos , arxiv:1412.5093 [ hep - th ] .
t. li , z. li and d. v. nanopoulos , arxiv:1502.05005 [ hep - ph ] .
t. li , z. li and d. v. nanopoulos , arxiv:1507.04687 [ hep - th ] . s. v. ketov and t. terada , phys .
b * 736 * ( 2014 ) 272 [ arxiv:1406.0252 [ hep - th ] ] .
s. v. ketov and t. terada , jhep * 1412 * ( 2014 ) 062 [ arxiv:1408.6524 [ hep - th ] ] .
t. terada , arxiv:1508.05335 [ hep - th ] .
j. r. ellis , c. kounnas and d. v. nanopoulos , phys .
b * 143 * ( 1984 ) 410 .
h. m. lee , jcap * 1008 * ( 2010 ) 003 [ arxiv:1005.2735 [ hep - ph ] ] .
r. kallosh and a. linde , jcap * 1306 * ( 2013 ) 028 [ arxiv:1306.3214 [ hep - th ] ] . l. j. hall and y. nomura , jhep * 1003 * ( 2010 ) 076 [ arxiv:0910.2235 [ hep - ph ] ] . l. j. hall and y. nomura ,
jhep * 1402 * ( 2014 ) 129 [ arxiv:1312.6695 [ hep - ph ] ] .
l. j. hall , y. nomura and s. shirai , jhep * 1406 * ( 2014 ) 137 [ arxiv:1403.8138 [ hep - ph ] ] .
l. e. ibanez , f. marchesano , d. regalado and i. valenzuela , jhep * 1207 * ( 2012 ) 195 [ arxiv:1206.2655 [ hep - ph ] ] .
l. e. ibanez and i. valenzuela , jhep * 1305 * ( 2013 ) 064 [ arxiv:1301.5167 [ hep - ph ] ] .
a. hebecker , a. k. knochel and t. weigand , jhep * 1206 * ( 2012 ) 093 [ arxiv:1204.2551 [ hep - th ] ] .
a. hebecker , a. k. knochel and t. weigand , nucl .
b * 874 * ( 2013 ) 1 [ arxiv:1304.2767 [ hep - th ] ] .
w. buchmuller , e. dudas , l. heurtier , a. westphal , c. wieck and m. w. winkler , jhep * 1504 * ( 2015 ) 058 [ arxiv:1501.05812 [ hep - th ] ] .
l. c. price , arxiv:1507.08360 [ astro-ph.co ] .
r. flauger , l. mcallister , e. pajer , a. westphal and g. xu , jcap * 1006 * ( 2010 ) 009 [ arxiv:0907.2916 [ hep - th ] ] . | large - field inflation in supergravity requires the approximate global symmetry needed to protect flatness of the scalar potential . in helical - phase inflation ,
the u(1 ) symmetry of the khler potential is assumed , the phase part of the complex scalar of a chiral superfield plays the role of inflaton , and the radial part is strongly stabilized .
the original model of helical phase inflation , proposed by li , li and nanopoulos ( lln ) , employs an extra ( stabilizer ) superfield .
we propose a more economical new class of the helical phase inflationary models without a stabilizer superfield . as the specific examples , the quadratic , the natural , and the starobinsky - type inflationary models
are studied in our approach .
+ revised version ut-15 - 33 + october 2015 desy 15 - 163 + ipmu 15 - 0143 3 cm * single - superfield helical - phase inflation * .3 in sergei v. ketov @xmath0 and takahiro terada @xmath1 .1 in _ @xmath2 department of physics , tokyo metropolitan university + minami - ohsawa 1 - 1 , hachioji - shi , tokyo 192 - 0397 , japan + @xmath3 kavli institute for the physics and mathematics of the universe ( ipmu ) + the university of tokyo , chiba 277 - 8568 , japan + @xmath4 institute of physics and technology , tomsk polytechnic university + 30 lenin ave . ,
tomsk 634050 , russian federation + @xmath5 department of physics ,
the university of tokyo , tokyo 113 - 0033 , japan + @xmath6 deutsches elektronen - synchrotron ( desy ) , 22607 hamburg , germany _
.1 in [email protected] , [email protected] .4 in |
three mode parametric instability in large scale , high optical power gravitational wave detectors was predicted by braginsky _
et al . _ in 2001 @xcite .
all subsequent analyses @xcite relied on the model prediction where amplitudes of certain acoustic modes of the interferometer mirrors would grow exponentially once an instability threshold of input laser power is reached .
it was generally considered that the exponential growth would eventually render the whole setup unstable and cause an interferometer to lose lock .
this prognosis , however , relied on a linearised approximation of the 3-mode optomechanical interaction which is valid only for small amplitudes of acoustic and stokes modes . for larger values , it is intuitively obvious that non - linearity should ultimately modify this growth . if the optical configuration can be maintained one would expect that the acoustic and higher - order optical oscillations should saturate .
knowing the amplitudes of such saturation effects as well as their timescale is of crucial importance for the operation of the real detectors now being implemented .
it is hard to overestimate the significance of rigorous analysis of this phenomenon in large - scale gravitational wave interferometers .
second generation detectors , such as advanced ligo @xcite are at the the latest stages of construction and testing .
these instruments are planned to have up to @xmath0 kw of circulating laser power in the arms .
as demonstrated in @xcite , the chance of 3-mode parametric instability at such high level of power is very high .
similarly , other advanced detectors , advanced virgo @xcite , kagra @xcite and geo - hf@xcite , might be susceptible to this effect , though with different probability ( see review @xcite for details )
. therefore , knowledge of the temporal dynamics and the values of final amplitudes the three participating modes reach at the saturation stage allows to design a feedback control system to suppress this instability before it develops .
several methods of mitigation this phenomenon were developed that can benefit from this information : ( i ) varying of the mirrors radii of curvature by heating @xcite ; ( ii ) decreasing acoustic modes q - factor @xcite ; ( iii ) introducing additional damping to acoustic modes via electrostatic feedback @xcite . in this paper
, we present a full non - linear treatment of this problem for the large - scale gravitational wave interferometers .
a similar approach has been used by polyakov and vyatchanin in @xcite to study the precursors of pi in the regime of input powers close to pi threshold . in this paper , we expand and generalise their treatment to arbitrary input power levels and investigate the time evolution of the amplitudes of stokes and pump optical modes as well as of acoustic mode .
noteworthy is the fact that the set of dynamical equations in this work is similar to those in @xcite , yet we prove that the requirement on smallness of the mechanical mode amplitude ( allegedly , it has to be smaller than the optical modes linewidth ) stated therein is not necessary , and this model is valid for arbitrarily large acoustic amplitudes . besides
, similar effect has been observed and analysed in small - scale whispering - gallery - mode optical resonators used to create stable radio - frequency optomechanical oscillators @xcite .
recently , the saturation of unstable oscillations has been observed at uwa and in lkb in tabletop optomechanical experiment using a high - finesse fabry - perot cavity with a silicon nitride membrane acting as the acoustic resonator test mass @xcite . in that paper , we derived a similar model to back the experimental results . in this paper , we focus on a different physical system that comprises massive freely suspended fabry - prot cavities which interact with high - power optical field circulating inside .
we also analyse the 3-mode parametric instability phenomenon in more detail and with greater generality .
the physics of three mode parametric instability is illustrated in fig .
[ fig : pi_dynamics ] .
we consider a fabry - perot interferometer pumped at a laser frequency @xmath1 close to the resonance frequency , @xmath2 , of one of the fundamental modes @xcite .
ultrasonic vibrations of the mirrors with frequency @xmath3 cause intracavity light to scatter into two motional sidebands , stokes , with frequency @xmath4 , and anti - stokes one , with frequency @xmath5 .
as the fundamental mode linewidth @xmath6 is normally much smaller than @xmath3 , the amplitude of motional sidebands is normally not enhanced by the cavity .
not so , however , if the sideband frequency coincides with one of the higher order optical mode ( hom ) frequencies .
such modes exist in any fabry - perot cavity and populate densely the free spectral range between the fundamental modes ( see sec .
3.3 of @xcite ) .
if the transverse spatial profile of such hom matches the profile of an acoustic vibration , the photons scattered into this mode from the fundamental mode build up , thereby channeling part of the optical power circulating in the fundamental mode to the hom .
the beat note of the fundamental mode and the hom , creates a near - resonant radiation pressure force on the mirror at the frequency of acoustic mode , @xmath3 .
now , there are two possibilities to consider .
if hom frequency coincides ( approximately ) with anti - stokes sideband frequency , @xmath7 , the radiation pressure force will be applied out of phase with the acoustic vibrations , thereby damping them @xcite .
this effect is analogous to radiation pressure cooling @xcite , where the scattered stokes photon energy is a sum of the fundamental mode energy and the acoustic mode phonon energy , _ i.e. _ @xmath8 .
the instability we are studying in this paper , on the contrary , occurs when the hom frequency matches the stokes sideband frequency @xmath9 : @xmath10 . to distinguish this hom from others
we will call it hereinafter a stokes mode . in this case ,
radiation pressure force of the beat note is in phase with the acoustic oscillations , leading to amplification .
this , in turn , makes amplitude of the stokes sideband larger , thereby increasing the amplitude of the radiation pressure force .
so , the loop closes and the instability breaks out . an illuminating description of this process in terms of feedback control theory can be found in the work of evans _
_ @xcite .
the picture above gives no account for natural decay of the stokes mode and acoustic oscillations due to loss , characterised by stokes mode linewidth @xmath11 stokes and acoustic mode decay rate @xmath12 .
these loss mechanisms counterbalance the instability , and for circulating power below the certain threshold ( see below ) , there is no instability .
however , as soon as the power circulating in the fundamental mode , reaches this threshold value , the 3-mode instability sets on .
very similar process happens in optical parametric oscillators , where above certain threshold pump power , in the presence of strong kerr non - linearity of the medium , pump photons scatter into pairs of _ idler _ and _ signal _ photons @xcite .
this process resembles a relief valve operation , when above certain pressure the valve opens and redirects the excess fluid flow into a reserve pipe , keeping the main pipe pressure constant .
similarly , above the instability threshold , all the excess optical power is redirected from the fundamental mode to the stokes mode and the acoustic mode oscillations .
new steady state amplitudes of all three participating modes are reached when the balance of power is restored in the system , meaning that the amount of power pumped into the system matches the sum power leaving it through the three decay channels , characterised by the decay rates of the two optical modes , @xmath6 and @xmath11 , and a mechanical decay rate , @xmath12 . in this paper , we present a non - linear theory of the 3-mode parametric instability in large scale gravitational wave interferometers with free suspended mirrors .
we derive and solve equations of motion for 3 participating modes amplitudes and obtain their temporal dynamics . to characterise the behaviour of unstable modes , we calculate values of new steady - state amplitudes and give an estimate of instability development timescale .
[ [ hamiltonian - of-3-mode - interaction . ] ] hamiltonian of 3-mode interaction .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + to represent parametric instability in large - scale gw interferometers , we use a simple model of a fabry - prot cavity with resonance frequency @xmath2 pumped with a laser , having frequency @xmath1 and power @xmath13 .
below , we choose to adhere to hamiltonian description of the system under study in contrast with the original work of braginsky _ et al . _
@xcite where a completely equivalent lagrangian approach has been used to derive equations of motion of interacting modes .
it is worth emphasising that our analysis below is purely classical and no quantum effects are taken into account in this manuscript .
nevertheless , the choice of hamiltonian classical description allows an easy expansion of this model to a quantum one . as a matter of fact
, it has been done recently in the work that predicted a new source of quantum radiation pressure noise in gravitational - wave interferometers originating from the 3-mode optomechanical interaction @xcite .
if we assume intracavity light to be linearly polarised and to propagate along the cavity optical axis @xmath14 , its electric and magnetic field strain components can be represented in terms of expansion over cavity modes as : @xmath15\,,\end{gathered}\ ] ] @xmath16\,,\end{gathered}\ ] ] where @xmath17 , @xmath18 , @xmath19 and @xmath20 stand for plank constant , speed of light and vacuum permittivity and permeability , respectively ; @xmath21 is the volume of the cavity of length @xmath22 occupied by a light beam with cross - section area @xmath23 ; @xmath24 is the j - th mode spatial distributions in the direction perpendicular to the propagation direction , @xmath25 denotes dimensionless complex amplitudes of the mode with frequency @xmath26 normalised so as @xmath27 represents the number of photons in the corresponding mode .
the acoustic mode of the mirror can be described in terms of surface deflection component along the z - axis : @xmath28 $ ] with @xmath29 being the transverse mode spatial shape and @xmath30 is the ground state oscillations amplitude for an oscillator with effective mass @xmath31 and eigenfrequency @xmath3 .
since parametric instability occurs only for the modes which satisfy a certain matching condition , @xmath32 ( detuning should be smaller than the larger of the stokes or fundamental mode line width , @xmath33 $ ] ) , we can limit our consideration to those three modes ( with @xmath34 ) .
the hamiltonian for this 3-mode interacting system reads : @xmath35\,,\end{gathered}\ ] ] with @xmath36 and the last term describing the well known optomechanical interaction when radiation pressure force ( @xmath37 light intensety ) acts on a mechanical degree of freedom .
after integration over transverse coordinates , @xmath38 , one can rewrite the it in a more familiar representation : @xmath39 where free evolution hamiltonians for optical modes can be written as @xmath40 ( @xmath34 ) , @xmath41 , optomechanical interaction terms read @xmath42 with optomechanical coupling strengths defined as @xmath43 where @xmath44 is an overlap factor of spatial profiles of 3 participating modes , defined as @xmath45 ^ 2\,,\ ] ] with @xmath46 . to complete a picture
we need to add a term responsible for coupling with the environment , @xmath47 that can be expressed in terms of mode decay rates , @xmath48 and @xmath12 , and corresponding external input fields , @xmath49 and @xmath50 : @xmath51 + i\hbar\sqrt{\gamma_m}[b_m^\dag b_{\rm th}+b_m b_{\rm th}^*]$ ] . here
@xmath52 includes external laser pumping amplitude @xmath53 and zero - mean fluctuations @xmath54 , while two other modes are driven by fluctuations only .
we assume optical mode fluctuations to be in vacuum state so as @xmath55 , and the mechanical damping noise , @xmath56 , corresponds to thermal white noise with correlation function @xmath57 , where @xmath58 is the average number of thermal phonons in the mechanical mode with @xmath59 boltzmann s constant and @xmath60 mode temperature ( usually , room temperature , @xmath61 k , is assumed ) . [
[ natural - scales - and - variable - renormalisation . ] ] natural scales and variable renormalisation .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + in this study , we are interested in classical large - amplitude dynamics of the system described by the above hamiltonian which means high occupation numbers for all participating modes .
therefore , we change from quantum units to more physically sensible ones within the frames of this problem . for optical modes , a steady state amplitude of light in the fixed length fabry - prot cavity , @xmath62
, looks a natural scale .
mechanical mode can be scaled by a displacement amplitude @xmath63 necessary to shift the stokes mode frequency by one half - linewidth , _
i.e. _ @xmath64 .
the latter is a natural non - linearity scale of an optomechanical system that sets the applicability limit for linearised model thereof .
hereinafter we will operate with scaled amplitudes defined as : @xmath65 similar transformation has to be applied to noise terms .
[ [ equations - of - motion . ] ] equations of motion .
+ + + + + + + + + + + + + + + + + + + + one can now write down heisenberg - langevin equations for the system described by hamiltonian in the frame rotating with each mode frequency .
this means doing substitutions @xmath66 and @xmath67 ( here we define frequency mismatch as @xmath68 ) , and then dropping all the terms oscillating with frequency @xmath3 and faster , one gets : [ eq : pieqns ] @xmath69 these equations can be rewritten in terms of dimensionless scaled amplitudes introduced in : [ eq : pieqns ] @xmath70 here we introduced an important quantity , a parametric instability ( pi ) gain , @xmath71 significance of which will be revealed below .
note that amplitude @xmath72 introduced by [ eq : normalisation ] practically coincides with dimensionless mechanical amplitude @xmath73 of @xcite .
as we are mostly interested in the strong signal dynamics of the 3-mode system , the noise terms in the above equations may be safely omitted . however , small initial nonzero amplitude of mechanical oscillations is necessary for nontrivial solution .
brownian thermal vibrations of the mirror provide this initial amplitude of @xmath74 which is equal to @xmath75 . solving this system of equations
numerically gives the characteristic result shown in fig .
[ fig : pi_dynamics ] .
[ [ parametric - instability - criterion . ] ] parametric instability criterion .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + for parametric instability to develop , the certain threshold of input pumping power has to be reached .
@xcite used the above introduced pi gain , @xmath76 as a figure of merit for pi in the resonance case of @xmath77 and defined it as : @xmath78 is the threshold input power value for resonant pumping .
if @xmath79 , the system goes unstable , otherwise no excitation of mechanical and stokes mode occurs . in a more general , detuned case
this criterion is only slightly modified : @xmath80 this condition as well as an instability onset time can be obtained using simple linearised model based on eqs . .
to start with we assume amplitude of a fundamental optical mode to be constant and equal to @xmath81 in the normalisation we chose . then from eqs
we obtain the following set of linear equations for the stokes and the mechanical modes : @xmath82 looking for general solution in the form @xmath83 , the pi condition , @xmath84 , is obtainable from the characteristic equation for the above linear system : @xmath85 with the solution : @xmath86\,,\label{eq : pigamma}\\ \nu & = \frac{\delta_m}{2}-\frac{1}{4}\sqrt{\sqrt{x^2+y^2}-x}\,,\label{eq : pinu}\end{aligned}\ ] ] where @xmath87 requirement @xmath88 yields the sought for relations and .
in gravitational - wave interferometers , as well as in small - scale , table - top optomechanical experiments @xcite , mechanical decay rate , @xmath12 , is way smaller than the decay rates of the optical degrees of freedom , @xmath89
. therefore one can safely assume optical modes to follow any changes in the mechanical mode almost instantaneously , without delay .
hence we can adiabatically eliminate optical modes by setting time derivatives in eqs . and to zero and express the two optical modes as functions of mechanical amplitude @xmath90 as : @xmath91 where we defined dimensionless detuning @xmath92 . substituting these expressions into the last eq .
, one arrives at a single non - linear differential equation for the mechanical amplitude @xmath72 : @xmath93 this equation displays the mechanism of saturation clearly .
indeed , the system gets unstable when the real part of the expression in brackets turns negative , meaning negative mechanical decay rate .
however , the rise of amplitude , @xmath72 , entering the denominator renders the negative term smaller and smaller , eventually reaching the critical point when the bracket turns 0 . from this
ensues a new , non - linear parametric instability condition of the form : @xmath94 where the non - linear gain , @xmath95 , gradually wanes as amplitude , @xmath72 , waxes , reaching a limit cycle .
the value of the critical amplitude is identical to the steady state amplitude , as @xmath90 is a monotone function of time and equals to : @xmath96^{1/2}\,.\ ] ] steady state amplitudes for optical modes immediately ensue from eqs and : @xmath97^{1/2}\,.\end{aligned}\ ] ] adiabatic limit can also help to understand the small hump one can notice on the plot of the stokes mode power in fig .
[ fig : pi_dynamics ] for pi gain @xmath98 which is absent on the other curves with lower gain .
the existence of this hump stems from the dependence of the stokes mode amplitude , @xmath99 , on the mechanical amplitude , @xmath72 , given by eq . .
if we consider the stokes mode normalised power @xmath100 and equate its time derivative to zero we get : @xmath101 which is a necessary condition for @xmath100 to reach its extremum . as we know from numerical solution , @xmath90 is monotonic , and @xmath102 means mechanical amplitude has reached its maximal steady state value @xmath103 .
therefore , the condition for the hump would be that the second term in the product above becomes equal to zero for @xmath104 that , accounting for the monotonic character of @xmath90 , means that the numerator of the second term is smaller than zero at @xmath105 . from this immediately ensues a condition on parametric gain value above which one shall observe this hump in temporal behaviour of the stokes mode : @xmath106 ^ 2+\delta_m^2\underset{\delta_m=0 , \kappa_0=\kappa_s } { \longrightarrow}4 \,.\ ] ] remarkably , in the resonance case , the value of pi gain @xmath107 , above which the hump develops in the stokes mode corresponds to the situation when the power circulating in the stokes mode gets equal to the threshold value , _
i.e. _ to the power circulating in the fundamental optical mode .
the hump represents the transient process of the initial excess growth of the stokes mode occupation number above the threshold value before the acoustic mode could reach its steady state occupation number ( recall that @xmath108 ) and further release of this level through inverse scattering to the carrier mode .
the latter process invokes recombination of the excess @xmath9 photons with acoustic phonons at @xmath3 yielding generation of @xmath2 photons . as shown by greyed out region in fig .
[ fig : pi_dynamics ] , when @xmath98 , the decrease of the fundamental mode power and the growth of the acoustic mode amplitude near the time when the stokes mode maximum is reached gets slower which is indicative of the intensification of inverse scattering .
[eq : p_m_ss ] ) , using parameters from table [ tab : params ] .
inset plot compares the behaviour of circulating power with ( thick red line ) and without ( thin red line ) pi .
_ lower panel : _ dependence of acoustic mode amplitude on input laser power for parameters listed in table [ tab : params].,title="fig : " ] + [eq : p_m_ss ] ) , using parameters from table [ tab : params ] .
inset plot compares the behaviour of circulating power with ( thick red line ) and without ( thin red line ) pi . _
lower panel : _ dependence of acoustic mode amplitude on input laser power for parameters listed in table [ tab : params].,title="fig : " ] [ [ approximate - solution . ] ] approximate solution .
+ + + + + + + + + + + + + + + + + + + + + by representing a complex function @xmath90 as @xmath109 one can easily obtain from the equations they satisfy : @xmath110 one can solve this system of first - order differential equations explicitly , but the result will be an implicit transcendental equation on @xmath72 that can not be resolved analytically .
nevertheless , a pretty good approximation thereto gives the solution of a simplified equation for @xmath72 , obtainable by expanding the bracket in the rhs of in @xmath111 around the @xmath103 of .
this results in a bernoulli equation @xcite of the shape : @xmath112 resolves in : @xmath113 .parameters used for simulation [ cols="<,^,^",options="header " , ]
it is obvious that in adiabatic limit ( @xmath114 ) , a sum of optical powers leaving the interferometer in the fundamental and the stokes modes , _ i.e. _ @xmath115 , must be equal to the power @xmath13 entering it .
power scattered in the acoustic mode is negligible compared to that of the optical modes due to a huge frequency difference . in order to see that it is indeed the case for our treatment one needs to write down standard input - output relations connecting light amplitudes outside the interferometer ( of incident and reflected beams ) with the intracavity ones derived in and :
@xmath116 where capital letters identify classical components of the light fields we are only concerned with in this work .
powers in the corresponding beams are related to these amplitudes as @xmath117 ( @xmath118 stands for @xmath119 ) .
assuming steady - state , _
i.e. _ setting @xmath120 in and , and normalisation relations , one gets expressions for the reflected light powers as follows : @xmath121 there is no problem to see now that @xmath122 and the power balance is observed for any value of the acoustic mode amplitude @xmath105 . substituting eq . in these equations ,
one gets the nonlinear i / o - relations for the 3-mode system above the pi threshold : @xmath123\,,\\ p^{out}_0 & = 4 p_{in } \dfrac{\sqrt{\mathcal{r}_0-\delta_m^2}-1}{\mathcal{r}_0}\,,\ \mathcal{r}_0 = \frac{p_{in}}{p_{in}^{\rm thr}}\,.\end{aligned}\ ] ] the dependence of outgoing power for each mode on pi gain and thereby on input power is drawn in fig .
[ fig : pout_vs_pin ] .
( red dash - dotted line ) and at the hom frequency ( blue line ) as a function of parametric gain @xmath76 and , thereby of input power .
note the non - linear and non - monotonic relation between input and reflected powers .
critical pi gain value of @xmath124 corresponds to the case of critical coupling , _
i.e. _ when the non - linear loss of carrier photons to the stokes mode and to the acoustic mode reaches the level of bare loss of the interferometer defined by the reflectivities of the mirrors.,scaledwidth=50.0% ] the interesting feature of these relation is their apparent non - linearity that is quite opposite to a naive assumption one might be tempted to make that intracavity circulating power in both optical modes is linearly proportional to the outgoing power .
note also the existence of critical value of input power corresponding to pi gain of @xmath107 ( in resonance case ) when the reflected power in carrier mode vanishes and the stokes mode output reaches its maximum .
physics of this process is straightforward , for it is this level of power when the loss of carrier photons due to 3-mode scattering ( into stokes and acoustic modes ) reaches the level of the cavity bare loss , summarised in @xmath6 .
this is a well known phenomenon of critical coupling .
this , in particular , has profound observational consequences , as measuring the pi by recording the beat note of the reflected lights at the carrier and stokes mode frequencies may result in zero signal for the range of pi gains around the critical one , _
i.e. _ around @xmath125 . returning to intracavity fields and recalling the definition of normalised modes , one can rewrite expressions and in more physical terms as : @xmath126\,,\label{eq : p_0_ss}\\ \dfrac{p_s}{\omega_s } & = \dfrac{p^{\rm thr}_0}{\omega_0}\left[\sqrt{\frac{p_{in}}{p_{in}^{\rm thr } } -\frac{4\delta_m^2}{\kappa_s^2}}-1\right]\,,\label{eq : p_s_ss}\\ \dfrac{p_m}{\omega_m } & = \dfrac{p^{\rm thr}_0}{\omega_0}\left[\sqrt{\frac{p_{in}}{p_{in}^{\rm thr } } -\frac{4\delta_m^2}{\kappa_s^2}}-1\right]\ , , \label{eq : p_m_ss}\end{aligned}\ ] ] where @xmath127 is the power circulating in the fundamental mode at resonance frequency @xmath2 when pi threshold is reached
. the plots of these expressions are shown in fig .
[ fig : pi_ss_amps ] .
here mechanical amplitude is inferred from the acoustic power using its relation to acoustic amplitude : @xmath128 , yielding : @xmath129\,.\ ] ] the above equations and have another interesting implication , namely the equality @xmath130 represents the well known manley - rowe relations for a non - linear interacting systems @xcite . in our case
it says that the number of stokes photons produced from the @xmath2 photons matches exactly the number of acoustic phonons generated in this process .
one might wonder if there is an inverse process going on in the system , _
i.e. _ the generation of @xmath2 photons from the pairs of @xmath9 photons and @xmath3 phonons , and if this process shall prevent the power circulating in the stokes mode to exceed that in the fundamental mode . indeed , such inverse scattering may happen and , were it not for constant pumping of laser photons into the system at frequency @xmath2 , there will be equal probability for direct and inverse scattering processes leading to equilibrium between the occupation numbers of the modes .
one has to remember about losses that constantly drain photons and phonons from the corresponding modes .
it is these losses that the pi threshold owes its existence to .
the threshold of pi represents the level of intracavity power , whereat no more @xmath2 photons can be born by the fundamental mode .
let the power in the stokes mode reach the level slightly higher than the threshold power as represented by the crossing of blue and red lines in the upper panel of fig .
[ fig : pi_ss_amps ] .
it means that input power is more than 4 times higher than @xmath131 .
the inverse scattering process creates then a photon at @xmath2 , thereby making fundamental mode to have one photon more than the threshold allows .
this photon can not decay away using the fundamental mode loss channel , as it is saturated at the 4 times lower input power level of @xmath131 .
the only way for it to escape is through scattering to the stokes photon and acoustic phonon again .
the probability of this scattering is higher than of escaping the cavity because the optoacoustic photon - phonon exchange rate @xmath132 is faster than the cavity decay rate @xmath6 , which is the prerequisite for parametric instability to start in the first place .
using expression for @xmath133 , one can derive a timescale for the instability onset , which is a timescale whereat linearised model breaks down and exponential ring - up gives place to a saturation and , eventually , to a new equilibrium state reached by a system .
it can be estimated as a moment when exponentially growing mechanical amplitude , @xmath134 , with @xmath135 , reaches the above calculated steady state level , @xmath103 of eq . , _
i.e. _ : @xmath136 this expression can be simplified if we recall that in real interferometers @xmath137 and condition @xmath138 should be satisfied for pi to arise .
expanding eq . in taylor series in @xmath139 and @xmath140 ,
one gets : @xmath141\,.\ ] ] thus one can get the following approximate expression for pi onset time : @xmath142 the dependence of pi onset time vs. pump laser power @xmath13 ( and pi gain ) is plotted in fig .
[ fig : pi_onset_time ] for parameters characteristic for advanced ligo detector and given in table [ tab : params ] . for this plot ) . ]
the above results are obtained for a single fabry - prot cavity .
however , they are easily generalised to the case of a power and signal recycled interferometer like the one of advanced ligo .
such an interferometer , if perfectly symmetric , is equivalent to two effective fabry - prot interferometers with effective linewidths , @xmath143 , and detunings , @xmath144 , defined in terms of arm cavities parameters and power / signal recycling mirror reflectivity and phase shift .
graphically , this fact is illustrated in fig .
[ fig : scallaw ] .
this result is well known as `` scaling law '' and devised by chen and buonanno in their seminal article @xcite .
the effective fabry - prot cavities represent so - called _ symmetric _ and _ anti - symmetric _ optical modes , which are coupled to corresponding _
common _ and _ differential _ acoustic modes of the mirrors ( see related definitions in sec .
5.3 of @xcite ) .
as shown in @xcite , the above consideration of 3-mode parametric instability in fabry - prot applies to each of these modes with the following substitutions of parameters : @xmath145\,,\\ \kappa_s & \to\kappa_{s\pm } = \kappa_s\re\left[\dfrac{1-\rho_{p , s}e^{2i\phi^s_{p , s}}}{1+\rho^s_{p , s}e^{2i\phi_{p , s}}}\right]\,,\\ \delta_m & \to\delta_m + \delta_\pm = \delta_m + \kappa_0\im\left[\dfrac{1-\rho_{p , s}e^{2i\phi_{p , s}}}{1+\rho_{p , s}e^{2i\phi_{p , s}}}\right]\,,\end{aligned}\ ] ] where @xmath146 are ( amplitude ) reflectivities of power and signal recycling mirrors , respectively , ( with @xmath147 being more habitual power transmissivities thereof ) , and @xmath148 are propagation phase shifts light of frequency @xmath149 acquires propagating the distance @xmath150 from arm cavities input test masses ( itms ) to the prm and srm , respectively .
here `` + '' sign stands for values that refer to symmetric mode , and `` @xmath151 '' sign indicates those of anti - symmetric one .
one can notice that the new linewidths ratio @xmath152 may be different from the one for the simple fabry - prot , as phase shifts @xmath153 deviate from @xmath154 , _
i.e. _ the fact that fundamental mode is resonant in pr / sr cavity does not mean that the same is true for the stokes mode .
the difference is @xmath155 which might be significant for @xmath156 metres long pr / sr cavities planned for advanced ligo .
asymmetry in interferometer arms does not change the general conclusion of this section , for the dual recycled interferometer can still be represented as two independent fabry - prot interferometers , as shown in @xcite . in this case , however , normal modes of the asymmetric system are not pure _ symmetric _ and _ anti - symmetric _ combinations of fields in the arm cavities , but a general linear combination thereof ( cf . section 2.6 of @xcite )
in this work , we analysed , using full non - linear treatment , the dynamics of 3-mode optomechanical instability in large - scale gravitational wave interferometers with freely suspended optics .
it turns out that intrinsic non - linearity of the 3-mode interaction does not allow excited unstable optical and acoustic mode amplitudes to grow unbounded , rather it makes them saturate to the new steady state values .
these values are governed by 3 dimensionless parameters : parametric gain , @xmath76 , normalised frequency mismatch ( detuning ) , @xmath157 , and the ratio of optical linewidths @xmath158 .
therefore , our theory can be equally applied to simple fabry - prot interferometers and to complex dual recycled interferometers of the second generation gravitational wave detectors , such as advanced ligo @xcite , advanced virgo @xcite , kagra @xcite and geo - hf@xcite .
our analysis shows that the process of instability development is quite slow for large scale detectors , lasting for many relaxation times of the acoustic modes , which , due to very high q factors of these modes , amounts to hundreds to thousands of seconds .
such a long onset time allows for efficient control and mitigation of this type of instability .
moreover , for reasonable values of pi gain , @xmath159 , consistent with recent parametric instability modelling for advanced ligo interferometers @xcite , acoustic mode amplitudes should not exceed nanometre level .
basing on this point as well as on the fact that the small - scale membrane - in - the - middle interferometer in uwa where pi was observed experimentally in the non - linear regime @xcite did not loose lock , one can presume that the same would be the case for a large - scale gw detector .
however , there remains a high level of uncertainty pertaining to a greater complexity of an electronic control of a large interferometer .
the response of the electronic control system to a slump of circulating power in the arms when pi starts to develop deserves a separate study .
it should be also noted that our theory assumes fixed , time - independent values of mode frequencies and optomechanical coupling . in real interferometers with suspended optics mirrors
tend to move around slowly and a laser beam spot does not rest at a fixed position on a mirror surface .
this results in slow compared to acoustic frequencies modulation of the frequencies of the stokes modes and thereby of the mismatch parameter , @xmath160 , and of the pi gain @xmath161 and of modes overlap factor @xmath162 , leading to reduced chance of instability .
the effect of such modulation will be reported elsewhere @xcite .
authors are grateful to andrey matsko , haixing miao and huan yang for fruitful and elucidating discussions , as well as to all the members of the macroscopic quantum mechanics discussion group for helpful comments and suggestions on improvement of our manuscript .
we also thank anonymous referees for constructive critics and useful comments that , we hope , helped refine the presentation and quality of our manuscript substantially .
the work of d.g.b , j.l . and c.z .
is supported by the australian research council .
p. fritschel , second generation instruments for the laser interferometer gravitational - wave observatory ( ligo ) , in _ gravitational wave detection , proc .
spie _ , edited by m. cruise and p. saulson , volume 4856 , page 282 , bellingham , wa , 2003 , spie . | we present a study of three - mode parametric instability in large - scale gravitational - wave detectors .
previous work used a linearised model to study the onset of instability .
this paper presents a non - linear study of this phenomenon , which shows that the initial stage of exponential rise of the amplitudes of a higher order optical mode and the mechanical internal mode of the mirror is followed by a saturation phase , in which all three participating modes reach a new equilibrium state with constant oscillation amplitudes .
results suggest that stable operation of interferometers may be possible in the presence of such instabilities , thereby simplifying the task of suppression . |
given an homeomorphism @xmath0 of the torus @xmath1 which is in the homotopy class of the identity , and a lift @xmath2 of @xmath0 to the universal covering , one can associate to @xmath3 its rotation set @xmath4 , a convex and compact subset of the plane consisting of all the limit points of sequences of the form @xmath5 , with @xmath6 and @xmath7 .
this definition was introduced by misiurewicz and ziemian in @xcite as a generalization of the rotation number for circle homeomorphisms , and has proved to be a useful tool in the study of the dynamics of these homeomorphisms . in particular , when the rotation set has nonempty interior it has been shown that @xmath0 exhibits very rich dynamics , with an abundance of periodic points and positive topological entropy @xcite .
but this complex behavior is not restricted to rotation sets with nonempty interior : there are several examples of homeomorphisms with rich dynamical properties such that their rotation set is a singleton . in this work
we are concerned with the interplay between transitivity of @xmath0 , transitivity on the universal covering space and rotation sets .
our motivation in studying transitivity on the universal covering space is to understand when it is possible for the dynamical system to exhibit this form of extreme transitivity , where there exist trajectories that are not only dense on the surface , but also explore all possible loops and directions .
this is a strictly more stringent concept , as ergodic rotations of the @xmath8-torus do not lift to transitive homeomorphisms of the plane .
there have been some prior works concerning transitivity of the lifted dynamics of surface homeomorphisms . in @xcite
, boyland defines @xmath9-transitivity for a surface homeomorphism as the property of having an iterate which lifts to the universal abelian covering as a transitive homeomorphism , and he characterized @xmath9-transitivity for rel pseudo - anosov maps . in @xcite
it was proven that a @xmath10 diffeomorphism of @xmath11 homotopic to the identity and with positive entropy has a rotation set with nonempty interior if @xmath0 has a transitive lift to a suitable covering of @xmath11 . in @xcite and @xcite , homeomorphisms of the annulus with a transitive lift
were studied , and it was shown that if there are no fixed points in the boundary then the rotation set of the lift contains the origin in its interior . more relevant to this work is the result from @xcite , where it is shown that if @xmath0 is a homeomorphism of @xmath11 homotopic to the identity which has a transitive lift @xmath3 to @xmath12 , then the origin lies in the interior of @xmath4 .
our main result is the converse of this fact : [ th : transitivoemcima ] let @xmath0 be a transitive homeomorphism of @xmath11 homotopic to the identity , and let @xmath13 be a lift of @xmath0 to @xmath12 .
if @xmath14 belongs to the interior of @xmath4 then @xmath13 is transitive .
note that , since the fundamental group of @xmath11 is abelian , the notion of @xmath9-transitivity ( as defined in @xcite ) in the case of a homeomorphism @xmath0 of @xmath11 is equivalent to saying that some iterate of @xmath0 has a lift to @xmath12 which is transitive . as an immediate consequence of theorem [ th : transitivoemcima ] together with the main result of @xcite , we have a characterization of @xmath9-transitive homeomorphisms of @xmath11 in the homotopy class of the identity : a homeomorphism of @xmath11 in the homotopy class of the identity is @xmath9-transitive if and only if it has ( a lift with ) a rotation set with nonempty interior . the key starting point for the proof of theorem [ th : transitivoemcima ] is a recent result from @xcite that rules out the existence of ` unbounded ' periodic topological disks .
after this , the theorem is reduced to a technical result , theorem [ th : maintheorem ] , the proof of which is obtained from geometric and combinatorial arguments relying on a well - known theorem of franks about realization of rational rotation vectors by periodic points @xcite and some classic facts from brouwer theory .
the next section states these preliminary facts and introduces the necessary notation . in section 3 ,
theorem [ th : maintheorem ] is stated , and its proof is presented after using it to prove theorem [ th : transitivoemcima ] .
we denote by @xmath15 the universal covering of @xmath1 . given a point @xmath16 , we denote by @xmath17 the projection of @xmath18 onto the first coordinate , by @xmath19 its projection onto the second coordinate , and @xmath20 the following definitions will simplify the notation in the proofs ahead .
for @xmath21 and @xmath22 , define @xmath23 and @xmath24 given @xmath25 , we denote by @xmath26 the translation @xmath27 .
if @xmath28\to{\mathbb{r}}^2 $ ] is curve , we denote its image by @xmath29 $ ] .
we say that @xmath30 is a _ @xmath31-translation arc _ if it joins a point @xmath32 to its image @xmath33 , and @xmath29\cap t_v([\gamma])=\gamma(1)$ ] .
the following lemma is a direct consequence of lemma 3.1 of @xcite .
[ lm : disjunto1vezdisjuntosempre ] let @xmath25 be a nonzero vector and let @xmath34 be an arcwise connected set such that @xmath35 .
then @xmath36 for all @xmath37 with @xmath38 . and the next result follows from theorem 4.6 of @xcite .
[ lm : brouwerline ] let @xmath39 and let @xmath40 be an arcwise connected subset of @xmath12 such that @xmath35 , and @xmath41 a @xmath31-translation arc disjoint from @xmath40 .
then , either @xmath42=\emptyset \quad \text { or } \quad k\cap\bigcup_{i\in{\mathbb{n}}}t^{-i}_v[\alpha]=\emptyset.\ ] ] we will also use the next theorem from @xcite .
[ th : franks ] let @xmath3 be a lift to @xmath12 of a homeomorphism of @xmath11 homotopic to the identity . if @xmath43 and @xmath44 are such that @xmath45 lies in the interior of @xmath46 , then there is @xmath47 such that @xmath48 let us recall some terminology used in @xcite .
an open set @xmath49 is _ inessential _ if every loop contained in @xmath50 is homotopically trivial in @xmath11
. an arbitrary subset of @xmath11 is called inessential if it has an inessential neighborhood , and _ essential _ otherwise .
if @xmath51 is an inessential subset of @xmath11 , and @xmath50 is an open neighborhood of @xmath52 , then @xmath53 is a closed inessential set .
this implies that every homotopy class of loops of @xmath11 is represented by some loop contained in @xmath50 , and for this reason any subset of @xmath11 with an inessential complement is called a _ fully essential _ set .
if @xmath54 is a homeomorphism , we say that a point @xmath55 is an _ inessential point _ for @xmath0 if there exists @xmath56 such that the set @xmath57 is inessential .
the set of all inessential points of @xmath0 is denoted by @xmath58 .
any point in the set @xmath59 is called an _ essential point _ for @xmath0 .
it follows from the definitions that @xmath58 is open and invariant , while @xmath60 is closed and invariant .
given an open connected set @xmath61 , we denote by @xmath62 the diameter in @xmath12 of any connected component of @xmath63 .
if @xmath64 , then we say that @xmath50 is _
bounded_. suppose that @xmath65 is a homeomorphism homotopic to the identity .
we need the following [ th : kota ] if @xmath0 is nonwandering and has ( a lift with ) a rotation set with nonempty interior , then @xmath58 is a disjoint union of periodic simply connected sets which are bounded . in particular ,
when @xmath0 is transitive one may conclude @xmath58 is empty : [ lm : essentialisall ] suppose that @xmath0 is transitive and has ( a lift with ) a rotation set with nonempty interior .
then @xmath66 .
suppose for contradiction that @xmath58 is nonempty .
then there is some point @xmath67 with a dense orbit .
let @xmath68 be the connected component of @xmath58 that contains @xmath18 .
by lemma [ th : kota ] , we have that @xmath68 is a periodic simply connected open and bounded set . thus any connected component @xmath69 of @xmath70 is bounded .
let @xmath71 be such that @xmath72 and let @xmath3 be a lift of @xmath0 .
then @xmath73 for some @xmath74 , and so if @xmath75 we have that @xmath76 .
the set @xmath77 is then bounded @xmath78-invariant ( because @xmath78 commutes with @xmath3 ) , and therefore @xmath79 is also bounded and @xmath78-invariant .
in particular all points in @xmath79 have a bounded @xmath78-orbit .
since @xmath80 is the @xmath0-orbit of @xmath68 , which contains the ( dense ) orbit of @xmath18 , it follows that @xmath81 .
the boundedness of @xmath82 then implies that @xmath83 .
it follows from these facts that every point of @xmath12 has a bounded @xmath84-orbit , from which we conclude that @xmath85 .
but it follows from the definition of rotation set that @xmath86 ( see @xcite ) , contradicting the fact that @xmath46 has nonempty interior .
theorem [ th : transitivoemcima ] is a consequence of the following technical result .
[ th : maintheorem ] let @xmath87 be a homeomorphism homotopic to the identity , and @xmath13 a lift of @xmath0 such that @xmath14 belongs to the interior of @xmath88 let @xmath89 be an open connected set such that @xmath90 is inessential and @xmath91 is fully essential .
then , for every @xmath92 there exists @xmath93 such that @xmath94 .
let @xmath95 be two open sets of the plane .
since @xmath0 is transitive , there exists @xmath96 such that @xmath97 intersects @xmath98 which implies that there exists @xmath92 such that @xmath99 let @xmath100 and @xmath101 be such that @xmath102 .
note that , as @xmath103 , the set @xmath104 is inessential .
since @xmath0 is transitive and @xmath4 has interior , lemma [ lm : essentialisall ] implies that @xmath66 and , in particular , @xmath105 is an essential invariant open set and , since @xmath0 is nonwandering , the connected component of @xmath106 in @xmath107 is periodic . in @xmath11 , an essential connected periodic open set is either fully essential or contained in a periodic set homeomorphic to an essential annulus , in which case the rotation set is contained in a line segment .
since @xmath4 has nonempty interior , we conclude that @xmath107 is fully essential .
therefore @xmath108 satisfies that the hypotheses of theorem [ th : maintheorem ] .
thus there exists @xmath109 such that @xmath110 intersects @xmath111 but this implies that @xmath112 intersects @xmath113 completing the proof .
since @xmath91 is fully essential , its preimage by @xmath123 , which we denote by @xmath124 , is a connected open set invariant by @xmath125 translates .
therefore , given a point @xmath126 in @xmath127 there exist two connected arcs @xmath41 and @xmath128 in @xmath124 such that @xmath41 connects @xmath126 to @xmath129 and @xmath128 connects @xmath126 to @xmath130 let @xmath131,\quad \gamma_{\beta}=\bigcup_{i=-\infty}^{\infty } t^{i}_{(0,1)}[\beta]\ ] ] and note that , as @xmath124 is @xmath125 invariant , then all integer translates of @xmath132 and @xmath133 are contained in @xmath134 fix an integer @xmath135\cup[\beta]\}$ ] .
then , since @xmath136\}<r\ ] ] and @xmath137 it follows that @xmath138 has at least two connected components , one containing the semi - plane @xmath139 and another containing @xmath140 likewise , @xmath141 has at least two connected components , one containing @xmath142 and another one containing @xmath143 by our choice of @xmath155 ( at the beginning of this section ) , @xmath156 is in the interior of @xmath46 .
it follows from theorem [ th : franks ] that there is @xmath157 such that @xmath158 .
since we may choose @xmath159 ^ 2 $ ] ( by using an appropriate integer translation ) , the proposition follows .
by proposition [ pr : fundamentaldomainmeets ] , if @xmath50 is the connected component of @xmath121 that contains @xmath120 ^ 2,$ ] then @xmath163 intersects @xmath164 , and since @xmath50 is bounded this implies that @xmath165 intersects @xmath166 since @xmath167 , it follows that @xmath168 if @xmath106 is a point in this intersection , then there exist integers @xmath169 $ ] and @xmath170 with @xmath171 , @xmath172 such that @xmath106 belongs to both @xmath173 and @xmath174 setting @xmath175 and @xmath176 yields the result .
[ lm : inbetween ] let @xmath177 be a nonzero vector , and @xmath178 , @xmath179 be two arcwise connected subsets of @xmath12 such that @xmath180 for @xmath181 .
suppose there are integers @xmath182 with @xmath183 , @xmath184 such that @xmath185 and @xmath186 .
then @xmath178 intersects @xmath179 .
moreover , there exists a @xmath31-translation arc @xmath30 contained in @xmath187 and joining a point @xmath188 to @xmath189 .
let @xmath192\to k_2 $ ] be a simple arc satisfying @xmath193 and @xmath194 since @xmath195 $ ] and @xmath178 are compact and @xmath196 is not null , there exists an integer @xmath96 such that , if @xmath197 then @xmath198 is disjoint from @xmath195.$ ] let @xmath199 : \alpha(t)\in \bigcup_{n=0}^{n_0}t^{-n}_{v}(k_1)\},\\ s_1 = \min\{t\in[s_0,1 ] : \alpha(t)\in \bigcup_{n=1}^{n_0}t^n_{v}(k_1)\},\nonumber\end{aligned}\ ] ] and let @xmath200 and @xmath201 be integers such that @xmath202 and @xmath203 finally , let @xmath204)\cup t^{j_0}_{v}(k_1),\ ] ] which is a connected set .
we claim that @xmath205 and @xmath206 .
to prove our claim , assume that it does not hold .
then @xmath207 .
note that by construction , @xmath208 is disjoint from @xmath209 since @xmath210 , it follows from lemma [ lm : disjunto1vezdisjuntosempre ] that @xmath211 for any @xmath212 , and so @xmath213 is disjoint from @xmath214 ( because @xmath215 ) . from these facts and from the hypotheses
follows that @xmath216 is disjoint from @xmath217 , so again by lemma [ lm : disjunto1vezdisjuntosempre ] we have that @xmath216 is disjoint from @xmath218 for all @xmath219 . but
@xmath220 , and clearly @xmath221 intersects @xmath216 .
this contradiction shows that @xmath222 , i.e. @xmath205 and @xmath223 since @xmath224 , we have shown that @xmath178 intersects @xmath179 .
note that @xmath225 , and let @xmath226\to k_1 $ ] be a simple arc joining @xmath227 to @xmath228 .
since @xmath229\subset k_1 $ ] and @xmath230 it follows that @xmath229\cap t_v(\alpha([s_0,s_1)))= \emptyset = t_v([\beta])\cap \alpha([s_0,s_1))$ ] .
therefore , letting @xmath30 be the concatenation of @xmath128 with @xmath231}$ ] we conclude that @xmath30 is a @xmath31-translation arc joining @xmath232 to @xmath189 .
[ pr : pintasegmento ] let @xmath248 be integers with @xmath249 , and @xmath250 if both @xmath251 and @xmath252 belong to @xmath253 then @xmath254 for each integer @xmath255 $ ] .
likewise , if both @xmath256 and @xmath257 belong to @xmath253 then @xmath258 for all integers @xmath255 $ ] . note that if @xmath265 and @xmath274 is chosen as in the statement , proposition [ pr : propertiesofsetofindices ] ( 1 ) implies that @xmath275 let @xmath276 be integers such that @xmath277 .
define , for @xmath278 .
for @xmath279 we will define @xmath280 in @xmath281 and @xmath282 recursively satisfying the following properties : suppose we have already defined @xmath280 and @xmath286 consider the @xmath287 different points of the form @xmath288 with @xmath289 , where @xmath290 .
note that from the recursion hypothesis ( 1 ) follows that @xmath291 , so that @xmath292 in particular , @xmath293 for each @xmath294 , so there exists @xmath295 , satisfying @xmath296 . since @xmath287 is greater than the cardinality of @xmath161 , by the pigeonhole principle there exists @xmath297 such that @xmath298 .
define @xmath299 and @xmath300 .
since both @xmath301 and @xmath302 belong to @xmath303 , it follows from proposition [ pr : pintasegmento ] that @xmath304 whenever @xmath305 .
note also that @xmath306 , and that @xmath307 .
so , @xmath308 and @xmath309 satisfy properties ( 1 ) and ( 2 ) ( see figure [ fig:2 ] ) .
if @xmath347 then , by proposition [ pr : cadawtemumtransladado ] , for each @xmath348 there exists @xmath349 such that @xmath350 .
this implies that @xmath351 intersects @xmath352 for each @xmath353 .
since the definition of @xmath354 implies that , for every @xmath355 , @xmath356 we conclude that @xmath357 intersects @xmath358 .
this means that @xmath359 intersects @xmath360 .
since @xmath361 , it follows that @xmath362 so @xmath363 .
assume , by contradiction , that there exists a sequence @xmath370 such that @xmath371 and such that , @xmath372 for all @xmath353 .
then , by proposition [ pr : pintasegmento ] , we can not have that both @xmath373 and @xmath374 belong to @xmath375 simultaneously , and neither can @xmath376 and @xmath377 . thus we may assume , with no loss of generality , that for infinitely many values of @xmath326 neither @xmath374 nor @xmath377 belong to @xmath375 ( since the other cases are analogous ) .
this means that @xmath378 extracting a subsequence of @xmath379 we may ( and will ) assume that the latter holds for _ all _ @xmath353 .
let @xmath380 since @xmath381 it follows that @xmath179 intersects the open connected sets @xmath382 and @xmath383 and therefore lemma [ lm : inbetween ] ( with @xmath384 ) implies that there exists a @xmath385translation arc @xmath41 contained in @xmath386 joining a point @xmath387 to @xmath388 note that the assumption that @xmath389 is disjoint of @xmath390 for all @xmath391 implies that @xmath392 for all @xmath393 .
furthermore , since @xmath372 , the set @xmath394 is also disjoint from @xmath382 .
thus @xmath195\cap { \widehat}f^{n_i}(o)=\emptyset$ ] for all @xmath353 .
let @xmath398\to t_v(o)$ ] be an arc joining @xmath399 to @xmath400 , and define @xmath401),\,\ , \alpha^{-}=\bigcup_{i\in{\mathbb{n}}}t^{-i}_{(1,0)}([\alpha]),\,\ , \beta^{+}=\bigcup_{i\in{\mathbb{n}}}t^i_{(0,1)}([\beta]),\,\ , \beta^{-}=\bigcup_{i\in{\mathbb{n}}}t^{-i}_{(0,1)}([\beta]).\ ] ] since @xmath394 is disjoint from @xmath195 \cup [ \beta ] \cup [ \gamma]$ ] , it follows from lemma [ lm : brouwerline ] that @xmath394 is disjoint from at least one of the following 4 connected sets : @xmath402 , \
, f_2=\alpha^{+}\cup\beta^{-}\cup[\gamma],\ ] ] @xmath403 , \ , f_4=\alpha^{-}\cup\beta^{-}\cup[\gamma].\ ] ] in particular there is @xmath404 such that @xmath405 is disjoint from @xmath406 for infinitely many values of @xmath353 .
we assume that @xmath407 , since the other cases are analogous . extracting again a subsequence of @xmath379
, we may assume that @xmath408 is disjoint from @xmath409 for all @xmath353 .
as @xmath394 is connected and contained in @xmath418 , for each @xmath353 the set @xmath405 is disjoint from either @xmath419 or @xmath420 .
this contradicts proposition [ pr : intersquadrantes ] , completing the proof of the claim .
let @xmath263 be such that the conclusion of lemma [ lm : pintasubquadrado ] holds with @xmath427 .
using again lemma [ lm : pintasubquadrado ] but setting @xmath428 , and @xmath429 , there exists @xmath430 , @xmath431 , an integer @xmath432 $ ] , and @xmath433 such that @xmath434 to complete the proof of theorem [ th : maintheorem ] , suppose , suppose that @xmath444 for all @xmath365 .
claims [ cl : mainclaim ] and [ cl : pintaquadradoeoposto ] together imply that , for sufficiently large @xmath368 both @xmath445 and @xmath446 .
this means that @xmath447 and @xmath448 . letting @xmath449 and @xmath450
, it follows that @xmath451 and @xmath452 .
thus , lemma [ lm : inbetween ] ( with @xmath453 ) implies that @xmath178 intersects @xmath179 , i.e. @xmath454 .
this contradicts our assumption at the beginning of this paragraph , completing the proof . | let @xmath0 be a transitive homeomorphism of the two - dimensional torus in the homotopy class of the identity .
we show that a lift of @xmath0 to the universal covering is transitive if and only if the rotation set of the lift contains the origin in its interior . |
we report on the discovery of a circumbinary disk seen in silhouette against the bright nebular background of the orion nebula . using multicolor observations taken with the hubble space telescope as part of the hst treasury program on the orion nebula cluster ,
we show that source 124 - 132 , a photoevaporated disk @xmath6 north of the trapezium previously imaged in h@xmath7 @xcite , harbors a binary .
binaries represent a typical product of the gravitational collapse of cores with high angular momentum .
in fact , the majority of stars in star - forming regions are in binary or multiple systems ( see duchene 1999 ; monin et al . 2007 ; and for a review mathieu 2000 and references therein ) .
also the onc contains a large number of binary stars @xcite , although the current counts indicate that the binary frequency , especially at the low - mass end , is lower by a factor 2 to 5 than in star - forming t associations like taurus - auriga @xcite . whether the relative paucity of binary systems in the onc is due to the initial conditions of the cloud or to a feedback " effect , like the ejection of binary companions in the cluster core caused by close dynamic encounters , is still a subject of debate .
there can be three disks in a young binary system : two circumstellar disks and a circumbinary one @xcite , with a complex geometry due to the potential variety of alignments between the disks and orbital planes @xcite , and interacting through transfer of energy and angular momentum @xcite .
disks have been detected around many spectroscopic binaries .
direct imaging of disks around wide , well separated binary systems , either circumstellar or circumbinary , is much rarer , the two most notable examples being the circumbinary disks around gg tau and uy aur .
the data presented here are part of the large dataset ( 520 images ) of acs exposures obtained for the _ hst treasury program on the orion nebula cluster _ ( go-10246 ) .
we mapped with acs / wfc an area of about 450 square arcmin nearly centered on the trapezium cluster in five filters : f435w ( johnson @xmath8 , 420s ) ; f555w ( johnson @xmath9 , 385s ) ; f685n ( h@xmath7 + [ ] @xmath10 , 340s ) ; f775w ( cousins @xmath11 , 385s ) ; and f850lp ( @xmath2-band , 385s ) . due to the adopted dithering strategy , most of the field has been exposed two times so the total integration time is typically twice that listed .
the drizzled acs images have been visually inspected and a master catalog of 219 circumstellar disks has been compiled @xcite .
figure 1 is extracted from this last paper , which also provides the fits files in electronic form .
further details on the _ hst treasury program _ observing strategy and data products will be given elsewhere ( robberto et al .
2008 , in preparation ) .
figure [ fig1 ] shows acs / wfc images of source 124 - 132 .
@xcite have already illustrated the h@xmath7 morphology of this object , a bright proplyd with a well resolved dark disk nearly perpendicular to the vertex of the ionization front . from their h@xmath7 analysis
, they estimate a disk tilt angle of @xmath12 .
the disk shows some point - like h@xmath7 emission on the west ( right ) side , which they interpret as either emission from a bipolar reflection nebula or a microjet perpendicular to the major disk axis . a more careful look at the h@xmath7 image shows that the point - like source is not centered on the dark disk axis , but shifted approximately 2 pixels ( @xmath13 ) to the north , away from the ionization front .
since the brightness of the ionization front does not allow us to trace the full extension of the disk to the south , this asymmetry is probably real .
the situation becomes clearer when our other images are considered . at shorter wavelengths ( f435w and f555w )
, the point - like source remains visible to the west of the disk axes , suggesting that the silhouette disk is slightly tilted with the western face toward us .
more interestingly , at longer wavelengths ( f775w and f850lp ) , two point - like sources appear well resolved along the disk axes . in figure [ fig2 ]
we plot the counts measured through the central part of the disk ( column 50 of the images in fig .
[ fig1 ] ) . in order to remove the contribution of the sky background and of the proplyd
we have subtracted from each cut the counts measured two pixels to the west ( column 52 ) .
figure [ fig2 ] clearly shows the two well resolved point sources appearing at the same position in the f775w and f850lp filters .
the f658n cut reveals the extent of the disk and shows that the disk is darker and possibly wider at the location of source s. for the f658n filter we have also added a line , corresponding to column 51 , to show the location of the h@xmath7 source discussed by @xcite .
source n remains visible in the f555w and f435w filters , whereas source s is perhaps barely detectable only in the f555w filter . in figure [ fig2 ]
we also indicate the limits of the disk , showing that the disk extends further on the northern side . this may be due to the presence of the ionized rim , which prevents tracking the disk in the vicinity of the ionization front .
in any case , we estimate a disk diameter of approximately 10 pixels , i.e. , 05 ( @xmath14 au ) , whereas the projected separation of the two stars is @xmath43 pixels , corresponding to 60 au . in table [ tbl-2 ] we present the magnitudes of the two sources measured by integrating the counts in a 3 pixel diameter aperture ; the sky has been measured at a close offset position free from ionized emission .
we have applied the acs zero points ( vega photometric system ) and aperture corrections from @xcite , averaging the encircled energy reported in table 3 of sirianni et al . for aperture radii of 1 and 2 pixels .
being hst diffraction limited , the psf in the widest filters depends on the source color .
following @xcite , we adjust the aperture correction in the f850lp filter to account for the extremely red colors of our sources . by repeating the measures with subpixel shifts of the centroid
, we estimate an uncertainty of approximately 0.1 mag in the f775w and f850lp passbands , and 0.2 mag in the bluer filters due to the faintness of the source and the highly non - uniform background .
the most striking result is the similarity between the magnitudes and @xmath15 colors of the two sources , with both of them having f775w@xmath16f850lp = 1.5 .
our images reveal the presence of a dark silhouette disk harboring two faint , similar objects . in the f775w and f850lp filters the sources appear clearly point - like ,
well separated , and without any evidence of bipolar extended emission from either sides of the disk , as normally observed when edge - on disks obscure their central stars ( * ? ? ?
* and references therein ) .
we will thus assume that we are seeing the reddened central stars rather than their diffuse light scattered by the disk surface . to constrain the nature of the two sources , we plot their position in a color - magnitude diagram ( fig . [
fig : isochrone ] ) and ( for source n only ) in a color - color diagram ( fig .
[ fig : color_color ] ) with respect to the 1 myr pms isochrone of , calculated in our passbands using the nextgenii synthetic atmosphere models @xcite .
we show the stellar population of the onc , we also show the displacement from the isochrone caused by extinction and mass accretion , both combined with the synthetic stellar spectra before convolution with the filter passbands . for the extinction we use both the standard @xmath17 reddening law and the @xmath18 reddening law , considered more appropriate for the orion nebula @xcite .
the accretion spectrum is reproduced following the @xcite recipe , combining @xmath19 of a @xmath20 k black body , accounting for the optically thick emission from the photospheric region heated by the shock of accreting material , with @xmath21 of optically thin emission from a dense slab ( @xmath22 @xmath23 ) of photoionized gas also at @xmath24 k , accounting for the pre - shock region .
the accretion luminosity , @xmath25 , and the stellar bolometric luminosity , @xmath26 , are both calculated integrating the model spectra , with @xmath25 finally multiplied by a normalizing factor .
this approximate treatment is adequate for our purposes , as we are interested in the intensity in broad - band filters rather than in the details of the true accretion spectrum .
figures [ fig : isochrone ] and [ fig : color_color ] show that the observed magnitudes of source n can be reproduced by different combinations of @xmath27 , @xmath28 and stellar mass . to find the best combinations of parameters , we calculate a large 3-d grid of models covering the full range of viable @xmath27 , @xmath28 and stellar mass ,
derive the synthetic photometry and search for the best match with the data .
the results are summarized in table [ table : intersections ] and marked as stars in figures [ fig : isochrone ] and [ fig : color_color ] .
it turns out that for each reddening law , there are typically two intersection points . for each reddening law ,
the two intersections provide combinations of lower and higher mass , extinction and accretion , ( the _ low _ and _ high _ solutions ) .
all four solutions place source n in the brown dwarf regime . to test the robustness of this result and to find the optimal solution , we have assumed that the uncertainties are dominated by photometric errors and performed a monte carlo simulation .
we generated a set of 10,000 trial stars populating a gaussian error distribution for the three magnitudes ; we then iterated our procedure to find the corresponding mass , accretion luminosity and extinction .
figure [ fig : mass_accr_relation ] shows the results for the two choices of @xmath29 .
the two areas at the top and bottom of each figure correspond to the _ high _ and _ low _ solutions , respectively .
the height of the peaks shows that the _ high _ solution is more compatible with @xmath17 , vice versa if @xmath18 .
only if @xmath18 the strongly degenerate _
high _ solution may lead to masses above the hydrogen burning limit ( @xmath30 ) .
another clue to the nature of the sources may come from a comparison of our estimated mass accretion rate with the values normally found for similar stars . using eq .
( 2 ) of @xcite with the stellar parameters from our baraffe / nextgen models , we derive for @xmath31 the values reported in table [ table : intersections ] . a comparison with figure 8 of @xcite , or figure 2 of @xcite ,
shows that only our _ low _ solutions provide mass accretion rates similar to those found in low - mass bds , whereas the _ high _ solutions give values in excess by 2 orders of magnitudes . if we exclude the _ high _ values on this basis , the most consistent solution for source n is the _ low _ case , corresponding to a @xmath32 brown - dwarf accreting at a rather standard @xmath33 yr@xmath34 , with about @xmath35 and @xmath18 , indicative of grains larger than those typically found in the interstellar medium .
similar conclusions should hold for source s , given the nearly identical @xmath1 and @xmath2 magnitudes .
even if ir spectroscopy will be needed to firmly establish the nature of these objects , the possibility that the two sources have sub - stellar mass appears quite robust .
if confirmed , this would be the first direct observation of a young brown - dwarf binary still accreting in its circumbinary disk . whereas our images shows the presence of a large - scale circumbinary disk , the evidence for mass accretion points to to the presence of at least one circumstellar disk .
theory provides a consistent scenario , predicting that circumstellar and circumbinary disks emerge as the mature outcome of the evolution of the original circumstellar disk in which the binary system formed @xcite . due to the exchange of angular momentum from the binary to the disk
, an inner gap develops between the inner and outer disk regions .
material can flow through the gap along spiral arms , feeding the circumstellar disks and therefore sustaining mass accretion into the central stars .
a detailed comparison of our system with the extant theoretical models is hindered , however , by the fact that our circumbinary disk is being photoevaporated , the first time such a phenomenon is observed .
the influence of uv flux from an external source on a circumstellar disk has been studied only for single stars by @xcite , who have shown that the extra heating of a disk face produces an increase of the disk flaring angle , leading to photoevaporation and possibly a warping of the system .
this may explain why source s remains undetected in our bluest filters : the southern side of the disk is the one more directly exposed to the uv flux and the stronger photoevaporation may locally increase the optical depth on this side . assuming the sources are identical , a line - of - sight difference
@xmath36 would bring source s below our detection limit at the shortest wavelengths while being compatible with the magnitudes observed in the red filters .
the work of lr and nd at stsci was done under the auspices of the stsci summer student program .
the authors wish to thank the anonymous referee for prompt and extremely valuable comments .
artymowicz , p. , clarke , c. j. , lubow , s. h. , & pringle , j. e. 1991 , , 370 , l35 artymowicz , p. , & lubow , s. h. 1994 , , 421 , 651 baade , w. , & minkowski , r. 1937 , , 86 , 123 baraffe , i. , chabrier , g. , allard , f. , & hauschildt , p. h. 1998 , , 337 , 403 bate , m. r. , & bonnell , i. a. 1997 , , 285 , 33 calvet , n. & gullbring , e. 1998 , , 509 , 802 duchne , g. 1999 , , 341 , 547 dutrey , a. , guilloteau , s. , & simon , m. 1994 , , 286 , 149 duvert , g. , dutrey , a. , guilloteau , s. , menard , f. , schuster , k. , prato , l. , & simon , m. 1998 , , 332 , 867 gnther , r. , & kley , w. 2002 , , 387 , 550 hauschildt , p. h. , allard , f. , baron , e. 1999 , , 512 , 377 johnson , h. l. 1967 , , 150 , l39 khler , r. , petr - gotzens , m. g. , mccaughrean , m. j. , bouvier , j. , duchne , g. , quirrenbach , a. , & zinnnecker , h. 2006 , , 458 , 461 lin , d. n. c. , & papaloizou , j. c. b. 1993 , protostars and planets iii , 749 luhman , k. l. 2004 , , 614 , 398 luhman , k. l. , et al .
2007 , , 666 , 1219 mathieu , r. d. 2000 , pre - main - sequence binary stars " in encyclopedia of astronomy and astrophysics , edited by paul murdin , article 2384 .
bristol : institute of physics publishing , 2001 menten , k. m. , reid , m. j. , forbrich , j. , & brunthaler , a. 2007 , , 474 , 515 mohanty , s. , basri , g. , & jayawardhana r. 2005 , astron .
326 , 891 monin , j .- l . , clarke , c. j. , prato , l. , & mccabe , c. 2007 , protostars and planets v , 395 muzerolle , j. , hillenbrand , l. , calvet , n. , briceo , c. , hartmann , l. 2003 , , 592 , 266 padgett , d. l. , strom , s. e. , & ghez , a. 1997 , , 477 , 705 petr , m. g. , coude du foresto , v. , beckwith , s. v. w. , richichi , a. , & mccaughrean , m. j. 1998 , , 500 , 825 prosser , c. f. , stauffer , j. r. , hartmann , l. , soderblom , d. r. , jones , b. f. , werner , m. w. , & mccaughrean , m. j. 1994 , , 421 , 517 reipurth , b. , guimaraes , m. m. , connelley , m. s. , & bally , j. 2007 , , 134 , 2272 ricci , l. , robberto , m. , & soderblom , d. r. , 2008 , , in press robberto , m. , song , j. , mora carrillo , g. , beckwith , s. v. w. , makidon , r. b. , & panagia , n. 2004 , , 606 , 952 robberto , m. , beckwith , s. v. w. , & panagia , n. 2002 , , 578 , 897 sirianni , m. , et al.2005 , , 117 , 1049 smith , n. , bally , j. , licht , d. , & walawender , j. 2005 , , 129 , 382 cccccc _ low _ & @xmath37 & @xmath38 & @xmath39 & 1.76 & @xmath40 + & @xmath41 & @xmath42 & @xmath43 & 1.95 & @xmath44 + _ high _ & @xmath37 & @xmath45 & @xmath46 & @xmath47 & @xmath48 + & @xmath41 & @xmath49 & @xmath50 & @xmath51 & @xmath52 + | we have found a photoevaporated disk in the orion nebula that includes a wide binary .
hst / acs observations of the proplyd 124 - 132 show two point - like sources separated by @xmath0 , or about 60 au at the distance of orion .
the two sources have nearly identical @xmath1 and @xmath2 magnitudes .
we analyze the brightest component , source n , comparing the observed magnitudes with those predicted using a 1myr baraffe / nextgen isochrone with different accretion luminosities and extinctions .
we find that a low mass ( @xmath3 ) brown dwarf @xmath41 myr old with mass accretion rate @xmath5 , typical for objects of this mass , and about 2 magnitudes of visual extinction provides the best fit to the data .
this is the first observation of a circumbinary disk undergoing photoevaporation and , if confirmed by spectroscopic observations , the first direct detection of a wide substellar pair still accreting and enshrouded in its circumbinary disk . |
central mass concentrations ( cmcs ) are frequently found in galaxies of all types , such as barred , spiral , disk or elliptical galaxies . a few examples of cmcs are : \1 .
large condensations of molecular gas with scales of 0.1 @xmath6 2kpc and masses of @xmath7 @xmath8 are found in the central regions ( e.g. , ohta , et al . , 1996 ;
sakamoto et al . , 1999 ; regan et al . ,
they are particularly evident in barred galaxies , where they are believed to be created by bar - driven inflow ( e.g. , athanassoula , 1992 ; heller & shlosman , 1994 ) .
the contribution of gas in the galactic evolution is a very important parameter .
the reader can find interesting information about the role of the gas in the review of kormendy & kennicutt ( 2004 ) and references therein .
it is widely accepted that most , if not all , of the galaxies have in their cores super - massive black holes ( smbhs ) .
central supermassive black holes seem to be a ubiquitous component in spiral disk galaxies , as well as in elliptical galaxies .
typical masses are in the range of @xmath9 @xmath8 ( e.g. , young , 1980 , turner , 1991 ; magorrian et al . , 1998 ; ferrarese & merritt , 2000 ; gebhardt et al . , 2000
; mclure & dunlop , 2001 ; tremaine et al . , 2002 ; wada & norman , 2002 ; kawakatsu & umemura , 2002 ; hopkins , et al . , 2006 ; shen , et al . ,
the reader can find more details regarding smbhs in the review of kormendy & richstone , 1995 .
dense star clusters are found near the centers of many spiral galaxies ( carollo , 2004 ; walcher et al . , 2003 ) ; these are young , very compact ( a few to up to 20 pc ) , and relatively massive ( @xmath10 @xmath8 ) . for the purposes of this study ,
a cmc is a sufficiently large mass at the center of a galaxy that is likely to have a dynamical effect on the evolution of its host galaxy . in other words , we focus on the dynamical consequences of a dense and massive nucleus .
dynamic searches indicate the presence of massive objects in the central regions of most galaxies .
( e.g. , m31 , m87 and ngc 4594 ) .
the estimation regarding the mass of the central region range from @xmath11 up to about @xmath12 . improved ground - based instruments and especially the hubble space telescope ( h.s.t ) , have given interesting information about the central regions of the galaxies .
dynamical evidence for the presence of a massive object in the core of a galaxy , is based on the value of the mass - to - luminosity ratio @xmath13 .
if this ratio increases towards the galactic center to values that are several times larger than normal , then this strongly indicates that a massive object has been discovered ( kormendy & rihstone , 1995 ) .
rapid rotation is also a strong indication for the presence of a massive object in the core of a galaxy ( miyoshi et al . , 1995 ;
greenhill et al . , 1997 ;
rubin et al . , 1997 ; bertola et al . , 1998 ;
sofue et al . , 1999 ) .
the effect of a central mass concentration ( cmc ) on the phase space in global galactic models has already been studied extensively in several earlier research works .
hasan & norman ( 1990 ) have shown that a black hole or a cmc in a barred galaxy can dissolve the barred structure . moreover , the fundamental orbits of the system ( called b orbits ) , supporting the bar are orbits elongated in the direction of the bar . as the black hole s mass increases , an inner lindblad resonance ( ilr ) appears and moves outward .
thus , the b orbits disappear when the ilr reaches the end of the bar .
this effect is significantly increased , if the bar is thinner .
orbits that stay closer to the central mass , with smaller jacobi constants , can change their character from regular to chaotic more easily .
furthermore , the study of hasan et al . ( 1993 ) indicates that , as the mass of the cmc increases the stable regular orbits in the region where the bar potential competes in strength with the central mass potential , especially around the region of the ilr will be no longer present and their position will be occupied by chaotic orbits .
shen & sellwood ( 2004 ) have conducted a systematic study of the effects of cmc on bars , using high quality @xmath14-body simulations .
they have experimented with both strong and weak initial bars and a wide range of the physical parameters of the cmc , such as the final mass , the scale length and the mass growth time .
their outcomes suggest that , for a given mass compact cmcs ( such as super - massive black holes ) are more destructive to barred structures than are more diffuse ones ( such as molecular gas clouds in many galactic centers ) .
they have shown that the former are more efficient scatters of bars supporting regulars elongated orbits , that pass close to the galactic center and therefore , decrease the percentage of the the regular orbits and increase the area of the chaotic region in the phase space .
all the above outcomes strongly indicate that a massive nucleus in the central parts of a galaxy , is a very important parameter of the dynamical system . in several previous works
, we see that low angular momentum stars in galaxies with massive nuclei , are scattered to much higher scale heights , displaying chaotic motion ( see caranicolas , 1997 , karanis & caranicolas , 2001 ; caranicolas & zotos , 2010 ; zotos , 2011a , b ) .
all the above studies , not only describe the chaotic character of orbits of low angular momentum stars , but also provide the main reason which is responsible for this behavior .
the reason is the presence of the strong force , which comes from the massive nucleus which resides in the core of the galaxy .
the dynamical models , used in all previous papers were global models . in caranicolas ( 1997 )
, it was used the mass model introduced by carlberg & innanen ( 1987 ) , while in karanis & caranicolas ( 2001 ) they have used a logarithmic potential .
the most characteristic result , derived in all cases , was the linear dependence of the critical value of angular momentum @xmath3 and the mass of the nucleus @xmath4 .
from our point of view , the background of this series of paper , in combination of other orbit calculations , was that managed to find and present , both numerically and semi - theoretically , relationships connecting the involved physical parameters of the dynamical system , such as the angular momentum and the degree of chaos in disk galaxies .
recently , caranicolas & zotos ( 2011 ) used a model consisting of two components .
the first component was a logarithmic potential describing a prolate elliptical galaxy , while the second one was a potential of a dense spherical nucleus .
two different cases were studied .
the case when the nucleus is absent and the case were we have an active galaxy hosting a dense and massive nucleus .
the main outcome of this work is that low angular momentum stars on approaching a dense and massive nucleus are deflected to higher values of @xmath15 displaying chaotic motion .
this suggests that it is the strong vertical force near the dense nucleus that is the main responsible for this scattering , combined with the star s low angular momentum which allows the star to approach sufficiently enough to the dense nucleus in order to be deflected . in an earlier work ,
caranicolas & papadopoulos ( 2003 ) adopted a simple time dependent model in order to study the transition from regularity to chaos in a galaxy , when mass is transported exponentially from the disk to the galactic core forming a dense and massive nucleus .
the results of this research suggest that during the galactic evolution , low angular momentum stars change their orbital character from regular to chaotic . in the present work
, we will continue this study in more detail .
in particular , we shall deploy this gravitational galactic model , in order to investigate the change of the orbital nature for stars moving in the meridian @xmath0 plane .
we will prove using a systematic and detailed study , that stars do not only change their nature from regular to chaotic during the galactic evolution , but there are also stars which maintain their character and also a small portion of stars that alter their character from chaotic to regular . the purpose of this research work is : ( 1 ) to find the correlations between the involved parameters of the system , ( 2 ) to explain the numerically obtained results using theoretical arguments , ( 3 ) to compare the critical value of the angular momentum @xmath3 with the value of the circular angular momentum @xmath5 at different distances from the center of the galaxy and ( 4 ) to study the evolution of orbits when mass is transported exponentially from the disk to the nucleus .
the present article is organized as follows : in section 2 we present our gravitational dynamical model . in section 3
we provide a detailed analysis of the structure of our time independent dynamical system .
moreover , some interesting numerical results are also provided in the same section . in section 4
we present some semi - theoretical arguments , in order to support and explain our numerically obtained outcomes . in the next section
we follow the evolution of the orbits in the time dependent model , when mass is transported from the disk to the nucleus .
we conclude with section 6 , where the discussion and the conclusions of this research are presented .
a very convenient method in order to study the behavior of the orbits of the stars in a galaxy , is to represent the galactic system by using a dynamical model , having about the same dynamical properties as the galaxy . in the present article , we shall study the transition from regular to chaotic motion , in an axially symmetric galaxy described by the disk - halo potential @xmath16 with @xmath17 ^ 2 \ \ \ , \ ] ] where @xmath0 are the usual cylindrical coordinates . here
@xmath18 is the mass of the disk , @xmath19 is the scale length of the disk , @xmath20 corresponds to the disk scale height , while @xmath21 is the core radius scale length of the halo component .
the three components of the disk @xmath22 , represent the fractional portions of old disk ( population ii stellar objects ) , dark matter and young disk ( population i stellar objects ) , respectively ( see clutton - brock et al . , 1977 ;
carlberg & innanen , 1987 ; caranicolas & innanen , 1991 ) .
the symbol @xmath23 denotes summation over these three components and also should apply that @xmath24 . at this point
, we should explain the role of the dark matter halo , since it is the dominant mass at large radii , where stars are deflected to larger apocenter upon interacting with the dark halo component .
the @xmath25 component represents the dark matter . in the present research for all our orbital calculations
the distance from the galactic center does not exceeds 10kpc .
therefore , we do not study the nature of orbits at large radii at which the dark matter is the dominant mass .
to potential ( 1 ) we add a spherically symmetric nucleus @xmath26 where @xmath4 is the mass and @xmath27 represents the scale length of the nucleus .
the plummer sphere we use in order to increase the central mass , has been applied many times in the past , in order to study the effect of the introduction of a central mass component in a galaxy ( see hasan and norman , 1990 ; hasan et al . , 1993 ) .
thus , the total potential is @xmath28 in this research , we use a system of galactic units , where the unit of mass is @xmath29 , the unit of length is 1kpc and the unit of time is @xmath30 yr . the unit of the angular momentum is 10 km kpc / s . the velocity unit and the energy unit ( per unit mass ) are 10 km / s and 100 ( km / s)@xmath31 respectively , while @xmath32 is equal to unity . in these units , we use the values : @xmath33 , @xmath34 ( kpc ) , @xmath35 ( kpc ) , @xmath36 and @xmath37 kpc , while the mass and the scale length of the nucleus are treated as parameters .
the above numerical values of the constant dynamical quantities of the system , secure positive density everywhere and free of singularities .
0.1 cm the rotation curve of the galactic model when @xmath38 and @xmath39 , is shown as the black line in figure 1 . in the same plot ,
the red line is the contribution from the spherical nucleus , while the blue line is the contribution from the disk - halo component .
we observe , that at small distances from the galactic center @xmath40 kpc , dominates the contribution from the spherical nucleus , while at larger distances @xmath41 kpc the disk - halo contribution is the dominant factor
. it would be very useful to compute the galactic mass density @xmath42 derived from the total potential ( 4 ) , using the poisson s equation @xmath43 \phi_{tot}(r , z ) \ \ \ .\end{aligned}\ ] ] remember that due to the axial symmetry the third term in eq .
( 5 ) is zero .
figure 2a shows the contours @xmath44 , for our model .
the values of the contours are : ( 0.010 , 0.017 , 0.030 , 0.052 , 0.150 ) . in figure 2b
, we see a 3d plot of the values of the density @xmath42 on the @xmath45 plane .
we can observe the spherical nucleus at the central region of the galaxy and also the well formed disk structure .
this plot assures that the mass density has positive value everywhere . for large values of @xmath46 and @xmath15
the mass density varies like @xmath47 and @xmath48 respectively .
this means that the total mass @xmath49 , enclosed in a sphere of radius @xmath50 increases linearly with distance .
this explains why circular velocity profiles , shown in fig .
1 remains almost flat for large values of @xmath46 .
here we must point out , that our gravitational potential is truncated at @xmath51 kpc , otherwise the total mass of the galaxy modeled by this potential becomes infinite , which is obviously not physical ( see binney & tremaine , 2008 ) .
as the total potential @xmath52 is axially symmetric and the @xmath2 component of the angular momentum is conserved , the dynamical structure of the system , can be studied using the effective potential @xmath53 in order to study the motion in the meridian @xmath0 plane .
the equations of motion are @xmath54 where the dot indicates derivative with respect to the time .
the corresponding hamiltonian can be written as @xmath55 where @xmath56 and @xmath57 , are the momenta per unit mass , conjugate to @xmath46 and @xmath15 respectively , while @xmath58 is the numerical value of the hamiltonian , which is conserved .
equation ( 8) is an integral of motion , which indicates that the total energy of the test particle is conserved .
the hamiltonian ( 8) , also describes the motion in the @xmath0 plane , rotating at the angular velocity @xmath59 all the outcomes of the present work , are based on the numerical integration of the equations of motion ( 7 ) , which was made by means of a bulirsh - ster method in fortran 95 , with double precision in all subroutines .
the accuracy of the calculations was checked by the consistency of the energy integral ( 8) , which was conserved up to the fifteenth significant figure .
in this section , we shall investigate the regular or chaotic nature of the orbits in our dynamical system .
figure 3 shows the numerically found relationship between the critical value of the angular momentum @xmath3 and the mass of the nucleus @xmath4 , for three different values of the scale length of the nucleus @xmath27 .
the critical value of the angular momentum @xmath3 , is the maximum value of the angular momentum @xmath2 , for which , stars scattered to the halo display chaotic motion for a given value of the mass of the nucleus @xmath4 and the scale length @xmath27 , when all other parameters are kept constant .
orbits were started near @xmath60 , with @xmath61 , while the initial value of the @xmath62 is always found from the energy integral ( 8) . the value of @xmath63 is the maximal root of the equation @xmath64 which was obtained numerically .
the value of the energy integral is @xmath65 and remains constant during this section .
the particular value of the energy was chosen so that in all cases @xmath66 and also in order to keep the initial radial and vertical velocities less than 55 km / s .
our numerical experiments indicate that the relationship between @xmath3 and @xmath4 is nearly a straight line in every case ( different values of @xmath27 ) .
each line divides the @xmath67 $ ] plane in two parts .
orbits with initial values of the parameters @xmath2 and @xmath4 on the upper left hand side part of the diagram , including the line , are chaotic orbits , while orbits with starting conditions on the lower right part of the same diagram are regular orbits . note that , for small values of the the angular momentum , the slope of the line is different .
moreover , it is interesting to point out that the extent of the area regarding chaotic motion , increases as the value of @xmath27 becomes smaller , that is when we have a more dense nucleus .
we shall come to that point later in the next section .
0.1 cm 0.1 cm 0.1 cm 0.1 cm in order to study the regular or chaotic nature of the orbits , we shall use the classical method of the @xmath1 , @xmath68 poincar surface of section technique , where @xmath56 and @xmath57 are the radial and the vertical momenta respectively .
if we set @xmath69 in equation ( 8) , we obtain the limiting curve in the @xmath1 phase plane , which is the curve containing all the invariant curves , for a given value of the energy integral @xmath58 .
the limiting curve of the dynamical system corresponds to @xmath70 0.1 cm figure 4a shows the @xmath1 phase plane when @xmath71 and @xmath39 . as we see the majority of the phase plane is covered by chaotic orbits .
regular orbits are confined mainly near the central part of the phase plane .
there are also several smaller islands embedded in the chaotic sea , which produced by secondary resonances .
figure 4b shows the @xmath1 phase plane when @xmath72 and @xmath39 . in this case , where the nucleus is more massive
, the extent of the chaotic region is larger , while the resonant phenomena are confined .
moreover , we observe that we have an increase in the radial velocity of the stars near the center of the galaxy . in figure
5a we see the @xmath1 phase plane when @xmath73 and @xmath39 . here
things are very different , since the majority of phase plane is covered by regular orbits now .
nevertheless a considerable chaotic area is present but it is confined mainly in the outer parts of the phase plane .
furthermore , we observe that the entire phase plane contains many sets of islands , which correspond to resonant orbits of higher multiplicity . in figure 5b
one can see the @xmath1 phase plane when @xmath74 and @xmath39 .
the main difference from fig .
5a is that now the nucleus is more massive and the small chaotic area of fig .
5a has become a vast chaotic sea .
furthermore , many of the sets of islands of invariant curves shown in fig .
5a disappeared .
figure 6a shows the @xmath1 phase plane when @xmath75 and @xmath39 .
we observe that the extent of the chaotic orbits has been decreased drastically as the value of the angular momentum increases . on the contrary , from the @xmath1 phase plane shown in figure 6b when @xmath76 and @xmath39 , it is evident that the increase in mass of the nucleus leads to an increase of the chaotic region .
we shall come to that later in this section . in figure
7a we see the @xmath1 phase plane when @xmath77 and @xmath39 .
we observe that the entire phase plane is covered only by regular orbits .
this is logical because the value @xmath78 of the angular momentum is larger than the critical value @xmath79 ( see fig .
figure 7b shows the @xmath1 phase plane when @xmath80 and @xmath39 .
as expected the increase of the mass of the nucleus has introduced chaotic areas in the @xmath1 phase plane . in all @xmath1 phase planes the value of
the energy integral is the same and equals to @xmath65 .
we observe from figs .
4 - 7 , that the extent of the chaotic regions in the phase plane increases as the nucleus becomes more massive . on the other hand , as we increase the value of the angular momentum @xmath2 , the chaotic regions become smaller .
therefore , we conclude that the portion of chaotic orbits with low or high values of angular momentum , in disk galaxies is larger when a massive and dense nucleus is present . here we must notice that the whole area of the @xmath1 phase plane is reduced and becomes smaller as @xmath2 increases . the outermost solid curve in the @xmath1 phase planes shown in figs .
4 - 7 is the limiting curve .
figure 8a - h shows eight representative orbits of the dynamical system .
figure 8a shows a regular orbit of 1:1 resonance when @xmath81 .
the initial conditions are : @xmath82 .
the value of @xmath62 is found always from the energy integral ( 8) for all orbits , while the energy integral has a constant value , which is @xmath65 . in figure
8b we see a periodic orbit , which is characteristic of the 2:1 resonance . for this orbit
we have @xmath83 , while the initial conditions are : @xmath84 .
this orbit carries stars out of the galactic plane .
figure 8c shows a periodic orbit of the 2:3 resonance , when @xmath83 .
the initial conditions are : @xmath85 . in figure 8d one
observes a periodic orbit when @xmath83 , with initial conditions : @xmath86 .
figure 8e shows a periodic orbit which is characteristic of the 4:3 resonance . here
@xmath87 , while the initial conditions are : @xmath88 . in figure 8f a periodic orbit of the 3:5 resonance family is shown . in this case
we have @xmath83 and the initial conditions are : @xmath89 . figure
8 g depicts a complicated periodic orbit of higher multiplicity . here
@xmath83 , while the initial conditions are : @xmath90 . a chaotic orbit is shown in figure 8h , when @xmath91 and the initial conditions are : @xmath92 .
note that stars moving in chaotic orbit as this shown in fig .
8h , are scattered to the halo when approaching the massive nucleus .
all orbits were calculated for a time period of 100 time units .
we can observe that the periodic orbits shown in figs .
8c , 8e , 8f and 8 g stay relatively close to the galactic plane and therefore , the corresponding families of regular orbits , support the disk structure of the galaxy .
figure 9 shows the percentage of the area @xmath93 covered by chaotic orbits in the @xmath1 phase planes as a function of the mass of the nucleus @xmath4 , for four different values of the angular momentum @xmath2 .
we observe that the percentage @xmath93 increases as the mass of the nucleus @xmath4 increases in four cases .
we must point out that from this diagram we can extract very useful information .
in particular , low angular momentum stars , display strong chaotic behavior .
moreover , as we increase the value of the angular momentum , the increase of the chaotic percentage , as the nucleus gaining mass becoming more massive , is more intense .
it seems that when the nucleus is massive enough , @xmath94 , the value of the angular momentum can not affect the nature of the orbits due to the fact that @xmath4 is the dominant parameter now .
we see in fig .
9 that all four fitting curves , which correspond to the four different values of the @xmath2 , converge when @xmath94 . we must point out , that the percentage @xmath93 is calculated as follows : we choose 1000 orbits with different and random initial conditions @xmath95 in each phase plane and then divide the number of those who produce chaotic orbits to the total number of orbits . in figure 10
we see a plot of the average value of the lyapunov characteristic exponent - l.c.e ( see lichtenberg and lieberman , 1992 ) as a function of the mass of the nucleus @xmath4 , for four different values of the angular momentum @xmath2 .
we can observe that the @xmath96 increases linearly , as the mass of he nucleus @xmath4 increases in four cases .
furthermore , we see that the fitting curves are almost parallel lines .
we conclude , that the degree of chaos increases linearly as the mass of the nucleus increases .
in addition , low angular momentum stars display larger degree of chaoticity than high angular momentum stars for the same mass of the nucleus .
we shall try to explain these results in the next section . here
we must notice , that it is well known that the value of l.c.e is different in each chaotic component ( see saito & ichimura , 1979 ) .
as we have in all cases regular regions and only one unified chaotic region in each @xmath1 phase plane , we calculate the average value of l.c.e by taking 500 orbits with different and random initial conditions @xmath95 in the chaotic region in each case .
note that , all calculated l.c.es were different on the fifth decimal point in the same chaotic region .
in the present section , we shall present some semi - theoretical arguments , together with elementary numerical calculations , in order to explain the numerically obtained relationships , given in figs . 3 , 9 and 10 . in order to explain the results shown in fig .
3 , we use a mathematical analysis , similar to that used in zotos ( 2011a , b ) .
our numerical simulations indicate that , near the nucleus , the @xmath97 force component is dominant ( it is approximately 20 times grater than at @xmath98 kpc , for the same value of @xmath15 ) and in fact in nearly equal to that produced by the nucleus itself .
the total @xmath99 force when @xmath38 and @xmath39 , is shown as the black line in figure 11 . in the same plot ,
the red line is the contribution from the spherical nucleus , @xmath100 , while the blue line is the contribution from the disk - halo component , @xmath101 .
even when @xmath15 is of order of unity , about @xmath102 of the total @xmath99 force is due to the massive nucleus . only when @xmath103 kpc the @xmath101 component becomes the dominant vertical force
furthermore , it is the nuclear force that causes the scattering to the halo leading to chaotic motion .
in fact , on approaching the nucleus .
there is a change in the star s momentum in the @xmath15 direction , given by @xmath104 where @xmath105 is the mass of the star , @xmath106 is the total ( i.e. , nuclear ) average @xmath15 force , while @xmath107 is the duration of the encounter .
it was observed , that the star s deflection into the halo , proceeds in each case cumulatively , a little more with each successive pass by the nucleus , rather than with a single dramatic " encounter .
it is assumed , that the star is scattered off the galactic plane and goes to the halo , after @xmath108 @xmath109 encounters , when the total change in the momentum in the @xmath15 direction , is of order of @xmath110 , where @xmath111 , is the tangential velocity of the star near the nuclear region .
therefore , we shall assume that @xmath112 setting @xmath113 in eq .
( 13 ) we obtain @xmath114 initially , if @xmath115 the orbit is chaotic .
otherwise , @xmath116 and after successive encounters with the galactic center , @xmath117 will increase , until it reaches @xmath118
. then the orbit will be chaotic .
the time needed to reach the condition @xmath119 is @xmath120 , where @xmath121 is the sum of the encounter durations ( 14 ) .
let @xmath122 be the age of the galaxy , which corresponds here to the final time of numerical integrations .
if @xmath123 , then some stars do not have enough time to be scattered .
we conclude from eq .
( 15 ) , that if @xmath124 , where @xmath3 is the angular momentum computed with @xmath125 , then there is still exist non - chaotic motions .
whereas for @xmath126 , all stars may have time to deploy chaotic motions . for a given value of @xmath127 (
the particular value is irrelevant as soon as @xmath127 is of order of unity ) , the quantitative behavior of @xmath128 force is shown in fig .
11 . for small values of @xmath15
, @xmath97 behaves as a harmonic oscillator force , while for larger values of @xmath15 , falls as @xmath129 . expanding the potential of the nucleus around the point @xmath130 , the square of the frequency of the harmonic motion in the @xmath15 direction is @xmath131 . substituting the harmonic motion law in eq .
( 15 ) with @xmath132 , we find @xmath133 where , for convenience , we have written @xmath134 instead of @xmath121 . if we set @xmath135 @xmath136 into eq .
( 15 ) , we obtain @xmath137 equations ( 16 ) and ( 17 ) are very simplistic , but nevertheless , reproduce the observed linear relationship between @xmath3 and @xmath4 , which is shown in fig .
3 . relation ( 17 ) also explains the dependence of the slope on the scale length of nucleus @xmath27 .
the larger is the @xmath27 the higher is the slope .
this is because @xmath121 is observed to be the about the same for the same scale length of the nucleus . on the other hand ,
when @xmath3 is very small , the star goes much closer to the nucleus , so that relation ( 16 ) must be used .
note that for the nucleus we always have @xmath138 and because of the smaller mass of the nucleus @xmath139 .
therefore , a lower slope results for the line in the region of very low angular momenta or , what is equivalent , for very low nuclear masses ( see fig .
the @xmath140 nuclear force is @xmath141 figure 12 shows a plot of the nuclear @xmath100 force as a function of the mass of the nucleus @xmath4 , for three different values of the scale length @xmath27 . as the scattering of the stars occurs near the nucleus , it must be at @xmath142 and @xmath143 .
the particular values of @xmath46 and @xmath15 are irrelevant . in this case
, we choose : @xmath144 kpc and @xmath145 kpc .
we observe that @xmath100 increases linearly with @xmath4 , in all cases .
one can see , that the pattern between @xmath140 and the mass of the nucleus ( shown in fig .
12 ) , is similar to those between the chaotic percentage @xmath93 or the average value of the l.c.e and the mass of the nucleus @xmath4 ( shown in figs .
9 and 10 respectively ) , which have been obtained numerically . in both cases
numerical outcomes suggest that , as the value of the mass of the nucleus @xmath4 increases , the percentage and also the degree of chaos in the dynamical system increases as well .
numerically found relationships given in figs . 9 and 10 can be explained because , as we have said earlier , the @xmath140 force is the main contributor , to the chaotic motion observed in the @xmath1 phase planes of the dynamical system .
moreover , from fig .
12 one can conclude that when the scale length of the nucleus @xmath27 is smaller , that is when the nucleus is more dense , the @xmath140 force is stronger , for each particular value of the mass of the nucleus @xmath4 .
a large number of orbits were computed , using different staring distances from the center of the galaxy , with the same low values of the initial radial and vertical velocities as was mentioned before .
orbits starting at the same distance @xmath46 , are considered to belong to the same group .
we have chosen this method of approach because the alternative , a self - consistent treatment with the energy as an additional parameter , in a scattering setting and in a galactic potential , is beyond the scope of the present paper .
some of these ideas may indeed be incorporated into a future work .
our numerical results suggest that , for each group of orbits , there is a linear relationship between @xmath3 and @xmath4 , similar to that shown in fig .
it is also of interest to compare the critical angular momentum @xmath3 with the circular angular momentum @xmath5 .
the circular angular momentum is given by the equation @xmath146^{1/2 } \ \ \ .\ ] ] such a comparison , for five values of the mass of the nucleus , when @xmath39 kpc , is given in table 1 .
orbits start at three different distances from the galactic center and according to the above , belong to three groups of orbits .
it is important to observe , that the ratio @xmath147 increases with decreasing @xmath46 .
this means that more stars ( initially in the plane ) are leaving the galactic plane resulting from chaos in the central region , than for larger values of @xmath46 .
this phenomenon could be taken as a first step for the explanation of the existence of a chaotic bulge in most disk - shaped galaxies , or perhaps , the presence of chaotic orbits contaminating " some more conventional stellar dynamical bulge . .a comparison between @xmath3 and @xmath5 for different distances from the galactic center . [ cols="^,^,^,^,^,^,^,^,^,^ " , ] figure 14a - d shows the evolution of the lyapunov characteristic exponent of four different orbits , as the total mass distribution of the dynamical system changes with time , following the set of equations ( 20 ) .
for all orbits shown in fig .
14a - d , the initial value of the hamiltonian ( 8) is @xmath148 , while @xmath149 and @xmath150 . in figure
14a we can see the evolution of the lyapunov characteristic exponent for an orbit and for a time period of @xmath151 time units , as the galaxy evolves .
the initial conditions are : @xmath152 , while the value of @xmath62 is found from the hamiltonian ( 8) in all cases .
the value of the angular momentum is @xmath153 . when @xmath154 time units , the mass of the nucleus is @xmath155 and the evolution stops .
the value of the hamiltonian is now @xmath156 .
note that the l.c.e tends asymptotically to zero , which clearly indicates that this orbit starts as regular and remains regular always near the galactic plane , during the galactic evolution .
figure 14b is similar to fig .
this orbit has initial conditions : @xmath157 , while the value of the angular momentum is @xmath158 .
when @xmath154 time units , the final mass of the nucleus is @xmath155 and the mass transportation stops .
the value of the hamiltonian is now @xmath159 . in this case
, the profile of the l.c.e clearly indicates that this orbit starts as a chaotic orbit and remains chaotic , during the galactic evolution .
we can observe that as the mass of the nucleus increases , during the mass transportation , the value of the l.c.e also increases gradually .
figure 14c depicts the evolution of the l.c.e of an orbit with initial conditions : @xmath160 .
the value of the angular momentum is now @xmath161 . after a time interval of 1000 time units
the mass of the nucleus is @xmath155 and the galactic evolution stops .
the hamiltonian settled to the value @xmath162 .
the profile of the l.c.e given in fig .
14c indicates that this orbit stars as a chaotic orbit but after the mass transportation it becomes an ordered one . in figure 14d
we observe the evolution of the l.c.e of an orbit with initial conditions : @xmath163 , while the value of the angular momentum this time is @xmath78 . after a time interval of 1000 time units
the final value of the mass of the nucleus is @xmath155 and therefore , galactic evolution stops .
the hamiltonian settled to the value @xmath164 .
the profile of the l.c.e given in fig .
14d shows that this orbit stars as a regular orbit but after the mass transportation , it becomes a chaotic one .
therefore , it is evident that if mass transport was not present , the orbit this time would have remain regular . in this case , the presence of the massive nucleus in the core of galaxy , has changed the nature of the orbit , from regular to chaotic .
furthermore , one can say that we are dealing with the case of fast " chaos , because the corresponding value of the l.c.e is relatively large ( see for details in caranicolas , 1990 and references therein ) .
as one would expect , the orbital evolution of stars in an evolving galactic model and the transition from regularity to chaos , should strongly depends on the timescale of the evolution , determined by the @xmath165 parameter . in order to reveal the importance of the timescale
, we will explore several cases , in which the temporal evolution timescale is varied .
figure 15a shows the evolution of the l.c.e of the orbit with the same initial conditions as in fig .
14c , for two different values of the timescale @xmath165 .
the blue line stand for @xmath166 , while the red one for @xmath167 .
in both cases , the orbit starts as a chaotic but after the mass transportation ( @xmath168 time units for the orbit shown in blue line and @xmath169 time units for the orbit shown in red line ) the orbit becomes regular .
note that both lines seem to converge when @xmath170 time units .
similar in figure 15b , we can see the the evolution of the l.c.e of the orbit with the same initial conditions as in fig .
14d , for two different values of the timescale @xmath165 .
again , the blue line stand for @xmath166 , while the red one for @xmath167 . in both cases ,
the orbit starts as a regular but after the mass transportation ( @xmath168 time units for the orbit shown in blue line and @xmath169 time units for the orbit shown in red line ) the orbit alters its nature and becomes chaotic .
note once more , that the values of the l.c.e converge when @xmath169 time units .
one should be very careful because , for small time intervals , a regular orbit may look chaotic and vise versa .
this is due to the different values of the timescale @xmath165 , which determines how fast the mass from the disk is transported to the nucleus .
this phenomenon is observed only when the galactic evolution ( mass transportation ) takes place .
when the galactic evolution stops and the system becomes time independent , the nature of the orbits is the same , regardless the particular value of @xmath165 we used .
thus , we reach to the conclusion that the timescale itself , does not affect directly the nature of the orbits during the galactic evolution .
it only determines each time , the time interval which is needed for an orbit during the mass transportation in order to form its final character ( regular or chaotic ) . using the above procedure , we have computed and tested a large number of orbits ( approximately 3000 different initial conditions of orbits for three different values of the timescale @xmath165 ) , with different starting distances from the center of the galaxy , while keeping the corresponding radial and vertical velocities smaller than 30 km / s .
orbits starting at the same distance @xmath46 were considered to belong to the same group .
all numerical outcomes suggest that , for each group of orbits , a linear relationship exists , between the critical value of the angular momentum @xmath3 and the final mass of the nucleus @xmath171 .
our numerical results indicate that the character of the orbits can change either from regular to chaotic and vise versa or not change at all , as the mass is transported and the massive nucleus is developed in the central region of the disk galaxy .
table 2 shows the respective percentage of total tested orbits in each case .
the symbols @xmath172 , @xmath173 , @xmath174 and @xmath175 correspond to each kind of the alteration of the orbital nature ( i.e. @xmath172 corresponds to regular to chaotic etc ) . from table 2
, it is evident that the formation of a massive nucleus , leads the majority of the orbits of the system to alter their nature from regular to chaotic or to stay chaotic .
some orbits maintain their regular character but only a very small portion of the total tested orbits change their nature from chaotic to regular during the mass transportation .
this tiny portion of initial conditions belongs to resonant orbits of higher multiplicity which correspond to multiple small islands of invariant curves in the @xmath1 phase plane .
these resonant orbits appear when the nucleus reaches its final value , that is @xmath155 .
furthermore , we may conclude that the particular value of the timescale @xmath165 does not affect directly the galactic evolution . the dominant parameter which determines the nature of an orbit is the final mass of the nucleus .
the timescale @xmath165 , simply shows how fast the galactic nucleus would reach its final mass through the mass transportation . whether the nature of an orbit will change or not during the galactic evolution described by the set of equations ( 20 ) , strongly depends on the initial conditions @xmath176 of each orbit .
moreover , as the change of the value of the hamiltonian ( 8) in each case is practically negligible , @xmath177 , we can say that the orbits in each group can be considered as isoenergetic .
in the present paper , we used an axially symmetric galactic gravitational model with a disk - halo and a spherical nucleus , in order to investigate the transition from regular to chaotic motion for stars moving in the meridian @xmath0 plane .
we studied in detail , the transition from ordered to chaotic motion , in two different cases : the time independent model and the time evolving model , when mass is transported from the disk to central nuclear region . in both cases
, we explored all the available range regarding the values of the main involved parameters of the dynamical system and we obtained interesting results , regarding the correlations between the important physical quantities of the system . here , we have to point out which of the outcomes of the present study are novel .
in this research we have used a composite disk - halo potential with an additional spherically symmetric nucleus , represented by a plummer sphere .
the current paper is basically consists of two parts . in the first part which covers sections 3 and 4 , describes the time independent galactic model .
he have conducted a thorough , detailed and systematic study in order to connect the involved parameters of the dynamical system .
in particular we obtained the relationship which connects the critical value of the angular momentum with the mass of the nucleus .
moreover , we connected the mass of the nucleus with the chaotic percentage and also with the average value of the lyapunov characteristic exponent , for several values if the angular momentum .
all these numerically obtained relationships , have been reproduced and verified using a semi - theoretical approach .
the second part of the present paper , that is section 5 , describes the time dependent model . here
, we adopted the simple model of mass transportation proposed by caranicolas & papadopoulos ( 2003 ) , as a starting point .
we studied in much detail this model using three different values of the timescale @xmath165 of the galactic evolution . for each value of the @xmath165 parameter , we computed 1000 different and random initial conditions @xmath178 .
our calculations suggest that the majority of the orbits of the system alter their nature from regular to chaotic or remain chaotic .
some orbits maintain their regular character but only a very small portion of the total tested orbits change their nature from chaotic to regular during the mass transportation .
it was found , that the particular value of the timescale @xmath165 does not affect directly the galactic evolution .
it is only determines how fast the nucleus would reach its final value of the mass .
the most significant revelation is that the dominant parameter , which determines whether an orbit would change or not and how its orbital nature during the mass transportation , is the final value of the mass of the nucleus .
it is evident that there is a well defined transition from regularity to chaos in our galactic model , for a given mass of the disk - halo component .
this transition is nearly a straight line connecting the points in the @xmath179 $ ] plane .
the slope of this line increases with the scale length of the nucleus @xmath27 .
the physical interpretation of this result is that , a nucleus of the same mass but with greater concentration ( smaller @xmath27 ) , deflects stars with higher values of angular momentum .
it also evident that the above straight line relation can be derived and reproduced using elementary mathematical calculations .
for this derivation , we have assumed that the deflection into a vertical chaotic orbit , happens after @xmath108 encounters with correlated deflections .
this choice was made by observing the behavior of the orbits in our numerical simulations and also by taking into account the smoothness of the total @xmath97 force ( see fig .
11 ) . in order to keep things simple ,
we have kept some of the parameters of the system constant and studied the behavior of orbits varying only three basic parameters @xmath180 .
it was found that , there is a linear relationship between the average value of the l.c.e and the mass of the nucleus @xmath4 . on the contrary
, the relation between the chaotic percentage @xmath93 on the @xmath1 phase plane and the mass of the nucleus @xmath4 is not linear but rather exponential .
the numerically found outcomes were explained , using some semi - theoretical arguments .
moreover , we concluded that when the scale length of the nucleus @xmath27 is smaller , that is when the nucleus is more dense , the @xmath140 force is stronger , for each particular value of the mass of the nucleus @xmath4 and thus the chaoticity of the dynamical system increases .
a comparison of the critical angular momentum @xmath3 to the local circular angular momentum @xmath5 , reveals that the ratio @xmath147 increases as we approach the center of the galaxy . for a galaxy with a nearly flat rotation curve ,
this variation mainly reflects the variation of the galactocentric distance @xmath46 .
therefore , stars moving near the center can more easily be deflected to higher values of @xmath15 .
this could be connected with the presence of massive spherical nuclei in the central regions of disk galaxies and suggest that near such massive nuclei the nature of the orbits of the stars must be highly chaotic , i.e. , with isotropic velocity dispersions .
our numerical experiments , show that near the center , the total @xmath97 force is entirely due to the massive nucleus and behaves like that of a harmonic oscillator .
this is a strong indication , that under certain conditions , the harmonic oscillator approximation can provide reliable results regarding the motion in the central regions of real galaxies . today , we know that the addition of matter onto compact objects , such as massive and dense galactic nuclei , is a very complicated astrophysical mechanism , which it would be very difficult to be studied and explored , if we do not establish some basic assumptions in order to make it more maneuverable . in the present research
, we have studied the change in the nature of the orbits in a time - dependent model , with an exponential mass transport from the disk to the nuclear region .
we have adopted this simple model , without going into more details for two basic reasons .
the first reason is that as it was mentioned above , it would be very difficult to construct a model in order to explain such a complicated mechanism and the second reason is that we are not interested in the astrophysical structure of the galaxy , but only in its dynamical behavior .
here we have to clarify , that our simple time dependent model , may not necessarily reflect the actual astrophysical mechanism .
nevertheless , similar simple dynamical models have been used frequently in the past ( see caranicolas & papadopoulos , 2003 ; caranicolas & innanen , 2009 ) in order to study the dynamical behavior of stellar orbits in complicated galactic systems .
once more , we have to point out that our current treatment of the system is purely dynamical .
this is exactly the main scope of this article .
it is well known that such mechanisms are met in the central regions of active galaxies and they are responsible for the enormous luminosity of the quasars which are hosted in the cores of the galaxies ( see collin & zahn , 1999 and references therein ) .
our numerical calculations show that a linear relationship between the the critical value of the angular momentum @xmath3 and the final mass of the nucleus @xmath171 also exists .
note , that the slope of this line increases as the value of the scale length of the nucleus @xmath27 increases .
this fact indicates that a nucleus of the same mass but with larger concentration ( smaller value of the @xmath27 ) is able to deflect stars with greater values of the angular momentum .
therefore , one can say that the present results are similar to those obtained in zotos ( 2011b ) , where we have also connected the involved parameters of the system with the extent of the chaotic regions and the angular momentum .
in addition to the observed similarities , we have also observed some differences in the behavior of orbits , between this model and the model used by zotos ( 2011a ) .
the most important difference is that in our previous work , the stars moved towards the halo after several encounters with the nucleus . here stars can stay for long time periods ( this actually depends on the particular value of the timescale @xmath165 , that is on how fast mass is transported from the disk to the nucleus ) , before they scattered to the halo .
therefore , one must be very cautious because , for small time periods , a chaotic orbit may look regular .
it seems that , in this case , one is not able to use methods for fast detection of chaos , especially when small values of @xmath165 are used ( see karanis & caranicolas , 2002 and references therein ) .
going on step further into the physical interpretation of the outcomes of the present work , we can summarize the following : there is no doubt that a large number of galaxies show significant activity in their nuclear regions .
we adopted a simple model , in order to express this activity , which is described by a mass transportation from the disk to the nuclear region .
the novelty and the aim of this research , is to explore on how the dynamical properties of the galaxy , that is the behavior of orbits , have been affected by the mass transfer phenomenon and also to establish a relationship between two important physical quantities : the final mass of the nucleus and the star s angular momentum .
the physical conclusion is that in active galaxies , a large number of low angular momentum stars display chaotic motion .
furthermore , some orbits that may look regular , in fact must be chaotic when the transportation of the mass is slow .
the general conclusion from our work is that the dominant parameters , which determine the nature of the orbits in galaxies are the mass of the nucleus and the angular momentum .
the rest involved quantities do not affect directly the nature of the orbits .
in particular , as we have seen in the time dependent model , the timescale @xmath165 does not really affect the final character of the orbits .
it is in our future plans to try a different way of mass transportation ( i.e a linear ) . from the present outcomes we may foresee , until proven , that even a different kind of mass transportation would not affect directly the nature of the orbits , since as we have seen the dominant parameter is the final mass of the nucleus and not how ( how fast or with which way ) we reach this value .
_ i am much indebted to dr .
hagai perets for reviewing the manuscript and also for all his very useful suggestions and comments which improved significantly the quality and the clarity of the present work .
athanassoula , e. 1992 , mnras , 259 , 345 .
walcher , c. j. , hring , n. , bker , t. , rix , h. w. , van der marel , r. , gerssen , j. , ho , l. c. & shields , j. c. 2003 , in carnegie observatories astrophysics series , vol . 1 : _ coevolution of black holes and galaxies _ , ed .
l. c. ho ( pasadena : carnegie obs . ) . | in the present article , we use an axially symmetric galactic gravitational model with a disk - halo and a spherical nucleus , in order to investigate the transition from regular to chaotic motion for stars moving in the meridian @xmath0 plane .
we study in detail the transition from regular to chaotic motion , in two different cases : the time independent model and the time evolving model . in both cases , we explored all the available range regarding the values of the main involved parameters of the dynamical system . in the time dependent model ,
we follow the evolution of orbits as the galaxy develops a dense and massive nucleus in its core , as mass is transported exponentially from the disk to the galactic center .
we apply the classical method of the poincar @xmath1 phase plane , in order to distinguish between ordered and chaotic motion .
the lyapunov characteristic exponent is used , to make an estimation of the degree of chaos in our galactic model and also to help us to study the time dependent model .
in addition , we construct some numerical diagrams in which we present the correlations between the main parameters of our galactic model .
our numerical calculations indicate , that stars with values of angular momentum @xmath2 less than or equal to a critical value @xmath3 , moving near to the galactic plane , are scattered to the halo upon encountering the nuclear region and subsequently display chaotic motion .
a linear relationship exists between the critical value of the angular momentum @xmath3 and the mass of the nucleus @xmath4 .
furthermore , the extent of the chaotic region increases as the value of the mass of the nucleus increases .
moreover , our simulations indicate that the degree of chaos increases linearly , as the mass of the nucleus increases .
a comparison is made between the critical value @xmath3 and the circular angular momentum @xmath5 at different distances from the galactic center . in the time dependent model
, there are orbits that change their orbital character from regular to chaotic and vise versa and also orbits that maintain their character during the galactic evolution .
these results strongly indicate that the ordered or chaotic nature of orbits , depends on the presence of massive objects in the galactic cores of the galaxies .
our results suggest , that for disk galaxies with massive and prominent nuclei , the low angular momentum stars in the associated central regions of the galaxy , must be in predominantly chaotic orbits .
some theoretical arguments to support the numerically derived outcomes are presented .
comparison with similar previous works is also made .
= 1 galaxies : kinematics and dynamics ; |
the ray - singer conjecture has been formulated in the seminal paper of ray and singer @xcite and proved independently by cheeger @xcite and mller @xcite for unitary representations .
its importance stems from the fact that as in the atiyah - singer index theorem , it equates analytic with combinatorial quantities , the analytic ray - singer and the combinatorial reidemeister torsions . by construction ,
both the analytic ray - singer and the combinatorial reidemeister torsions provide canonical norms on the determinant line of cohomology .
there have been various approaches to obtain a canonical construction of analytic and reidemeister torsions as elements instead of norms of the determinant line of the cohomology .
these constructions seek to refine the notion of analytic and reidemeister torsion norms on that determinant line , which basically corresponds to fixing a complex phase in the family of complex vectors of length one . in case of the reidemeister torsion
this has been done by farber and turaev @xcite and @xcite .
refinement of analytic torsion has been studied by the first author jointly with kappeler in @xcite and @xcite , as well as by burghelea and haller in @xcite and @xcite .
both notions have subsequently been compared by the first author jointly with kappeler in @xcite .
an extension of refined analytic torsion to manifolds with boundary has been undertaken by the second author @xcite and lee and huang @xcite in two different independent constructions .
recently , lee and huang also compared the two notions of refined analytic torsion on manifolds with boundary in @xcite .
the fundamental property of the refined analytic and farber - turaev torsions is that they define analytic functions on the complex representation space .
the main result of the present discussion is an extension of this result to the refined analytic torsion on manifolds with boundary , introduced by the second author in @xcite . as a consequence we establish the gluing property of refined analytic torsion on connected components of the representation variety that contain a unitary point .
the gluing formula for refined analytic torsion may be used to prove a gluing result for the ray - singer torsion norm for certain non - unitary representations path - connected to a unitary element .
however we chose to devote the final two sections of the present paper to an alternative proof of the gluing property for the ray singer analytic torsion for non - unitary representations , which is stronger since we do not single out connected components without unitary elements .
the ray - singer theorem has been extended to unimodular representations by mller @xcite . in case of a general non - unitary representation ,
the quotient of the analytic and reidemeister torsion norms admits additional correction terms which have been studied by bismut and zhang in @xcite . in a separate discussion @xcite ,
the authors employ analyticity of refined analytic torsion to provide an alternative derivation of the bismut - zhang correction terms for the connected components of unitary points in the representation variety . both the analytic and combinatorial reidemeister torsions make sense on compact manifolds with boundary after posing relative or absolute boundary conditions .
the gluing property of the analytic torsion , which is foremost a spectral invariant , is striking and has been proved by lck @xcite , vishik @xcite and generalized by lesch @xcite , under the assumption of product metric structures and unitary representations .
the anomaly of analytic torsion on a compact manifold with boundary with a non - unitary representation and general metric structures near the boundary , has been studied by brning and ma in @xcite .
recently , brning and ma established in a follow - up paper @xcite the gluing formula for analytic torsion in case of a non - unitary representation , generalizing previous results in @xcite and @xcite .
we present an alternative proof of a result by brning - ma @xcite by a temporal gauge transformation argument .
this section reviews the construction by the second author @xcite
. + let @xmath0 be a compact oriented odd - dimensional riemannian manifold with boundary @xmath1 . consider a complex representation @xmath2 of the fundamental group @xmath3 on @xmath4 .
let @xmath5 be the induced flat complex vector bundle over @xmath6 with monodromy equal to @xmath2 and no canonical choice of @xmath7 in case @xmath2 is not unitary .
the flat covariant derivative @xmath8 acts on sections @xmath9 and extends by leibniz rule to a twisted differential on @xmath10-valued differential forms @xmath11 , where the lower index refers to compact support in the open interior of @xmath6 .
this defines the twisted de rham complex @xmath12 . the metrics @xmath13 induce an @xmath14-inner product on @xmath11 .
we denote the @xmath15completion of @xmath11 by @xmath16 . throughout this section
the representation @xmath2 is fixed and we omit the lower index @xmath2 in the notation of @xmath5 in most of the discussion .
next we introduce the notion of the dual covariant derivative @xmath17 .
it is defined by requiring for all @xmath18 and @xmath19 @xmath20=h^e(\d_xu , v)+h^e(u,\d'_xv).\end{aligned}\ ] ] in the special case that @xmath21 is unitary , the dual @xmath17 and the original covariant derivative @xmath22 coincide .
as before , the dual @xmath17 gives rise to a twisted de rham complex @xmath23 . for any differential operator
@xmath24 acting on @xmath25 , we denote by @xmath26 its minimal graph - closed extension in @xmath27 .
the maximal closed extension is defined by @xmath28 . by brning and lesch ( * ? ? ?
* lemma 3.1 ) , the extensions define hilbert complexes @xmath29 , where @xmath30 , and @xmath31 , where @xmath32 .
the laplace operators , associated to these hilbert complexes are respectively defined as @xmath33 similar definitions hold for the dual connection @xmath34 and for the laplace operators @xmath35 and @xmath36 of the hilbert complexes @xmath37 and @xmath38 respectively .
the difference @xmath39 is a bounded endomorphism valued operator and hence the equality of domains @xmath40 the following theorem , compare ( * ? ? ?
* theorem 3.2 ) , summarizes the classical de rham theorem on manifolds with boundary , cf .
* remark after proposition 4.2 ) and ( * ? ? ?
* theorem 4.1 ) ; strong ellipticity of the corresponding laplace operators follows from ( * ? ? ?
* lemma 1.11.1 ) .
[ thm41 ] the hilbert complexes @xmath29 and @xmath31 are fredholm and the associated laplacians @xmath41 and @xmath42 are strongly elliptic .
the cohomologies @xmath43 and @xmath44 of the fredholm complexes @xmath29 and @xmath31 , respectively , can be computed from the following smooth subcomplexes , @xmath45 respectively , where @xmath46 denotes the natural inclusion of the boundary .
corresponding statement holds also for the complexes associated to the dual connection @xmath17 .
the riemannian metric @xmath47 and a fixed orientation on @xmath6 define the hodge star operator @xmath48 and the chirality operator ( @xmath49 ) @xmath50 this operator extends to a self - adjoint involution on @xmath27 .
the following properties of @xmath51 are essential for the construction below , cf . @xcite .
[ g - lemma ] the self - adjoint involution @xmath51 on @xmath27 maps @xmath52 to @xmath53 , and @xmath54 to @xmath55 . with @xmath51
restricted to appropriate domains , we have @xmath56 [ domain ] we introduce the doubled hilbert complexes @xmath57 similar to , we have the equality of domains @xmath58 the self - adjoint involution @xmath51 gives rise to the `` chirality operator '' @xmath59 an immediate consequence of lemma [ g - lemma ] is the following the chirality operator @xmath60 acts as @xmath61 moreover we have the relation @xmath62 .
we now apply the concepts of @xcite to our new setup and define the odd - signature operator of the hilbert complex @xmath63 by @xmath64 by @xcite the odd signature operator @xmath65 is strongly elliptic with discrete spectrum and an agmon angle @xmath66 .
consider for any @xmath67 the spectral projection of @xmath68 onto eigenspaces with eigenvalues of absolute value in the interval @xmath69 $ ] : @xmath70}:=\frac{i}{2\pi}\int_{\gamma(\lambda)}(\b^2-x)^{-1}dx,\ ] ] with @xmath71 being a closed counterclockwise circle around the origin surrounding eigenvalues of absolute value in @xmath69 $ ] . by the analytic fredholm theorem
, the range of the projection lies in @xmath72 and the projection commutes with @xmath68 .
moreover , @xmath73}$ ] is of finite rank and the decomposition @xmath74}\oplus \textup{image}\,(\one - \pi_{\b^2 , [ 0,\lambda]}),\end{aligned}\ ] ] is a direct sum decomposition into closed subspaces of the hilbert space @xmath75 .
note that if @xmath21 is unitary and hence @xmath68 is self - adjoint , the projection @xmath76}$ ] is orthogonal . induces a decomposition of @xmath77 @xmath78}\oplus \domm_{(\lambda , \infty)}.\ ] ] since @xmath79 commutes with @xmath80 and hence also with @xmath73}$ ] , we obtain a decomposition of @xmath63 into subcomplexes @xmath81 } , \dd_{[0,\lambda]})\oplus ( \domm_{(\lambda , \infty ) } , \dd_{(\lambda , \infty ) } ) \\ & \textup{where } \ \dd_{\mathcal{i}}:=\dd|_{\domm_{\mathcal{i } } } \ \textup{for } \ \mathcal{i}=[0,\lambda ] \
\textup{or } \
( \lambda , \infty ) .
\end{split}\ ] ] the chirality operator @xmath60 commutes with @xmath80 and respects the decomposition so that @xmath82 } \oplus \gg_{(\lambda , \infty ) } , \quad \b=\b^{[0,\lambda]}\oplus \b^{(\lambda , \infty)}.\ ] ] [ bijective][cohomology ] ( * ? ? ?
* corollary 3.14 and 3.15 ) . the operator @xmath83 , @xmath67 is bijective .
the complex @xmath84 is acyclic and @xmath85 } , \dd_{[0,\lambda]})\cong h^*(\domm , \dd).\ ] ] recall the notion of a determinant line of a finite dimensional complex @xmath86 @xmath87 where @xmath88 is the top exterior power of @xmath89 and the @xmath90 upper index denotes the dual vector space .
we follow ( * ? ? ?
* section 1.1 ) and define the _ refined torsion element _ of the complex @xmath91 } , \dd_{[0,\lambda]})$ ] @xmath92}:=c_0\otimes ( c_1)^{-1}\otimes \cdots \otimes ( c_r)^{(-1)^r } \otimes ( \gg_{[0,\lambda]}c_r)^{(-1)^{r+1}}\otimes \cdots & \\
\cdots \otimes ( \gg_{[0,\lambda]}c_1 ) \otimes ( \gg_{[0,\lambda]}c_0)^{(-1 ) } \in \textup{det}(h^*(\domm_{[0,\lambda ] } , \dd_{[0,\lambda ] } ) ) , & \end{split}\ ] ] where @xmath93 } , \dd_{[0,\lambda]})$ ] are arbitrary elements of the determinant lines , we denote the extension of @xmath94}$ ] to a mapping on determinant lines by the same letter , and for any @xmath95 } , \dd_{[0,\lambda]})$ ] the dual @xmath96 } , \dd_{[0,\lambda]})^{-1}\equiv \det h^k(\domm_{[0,\lambda ] } , \dd_{[0,\lambda]})^*$ ] is the unique element such that @xmath97 . by proposition [ cohomology ] we can view @xmath98}$ ] canonically as an element of @xmath99 , which we do henceforth .
the fundamental part of the construction is the graded determinant .
the operator @xmath100 is bijective by proposition [ bijective ] and hence by injectivity ( put @xmath101 to simplify the notation ) @xmath102 moreover the complex @xmath103 is acyclic by proposition [ cohomology ] and due to @xmath104 being an involution on @xmath105})$ ] we have @xmath106 we have @xmath107 and by surjectivity of @xmath108 we obtain from the last three relations above @xmath109})=\textup{ker}(\dd_{\mathcal{i}}\gg_{\mathcal{i } } ) \oplus\textup{ker}(\gg_{\mathcal{i}}\dd_{\mathcal{i}}).\end{aligned}\ ] ] note that @xmath65 leaves @xmath110 and @xmath111 invariant .
hence , we put @xmath112 we arrive at a direct sum decomposition @xmath113 by @xcite , there exists an agmon angle @xmath114 for @xmath65 , which is clearly an agmon angle for the restrictions above , as well . for strongly elliptic boundary value problems @xmath115 of order @xmath116 on @xmath6 with an agmon angle @xmath66 , the associated zeta - function
is defined by @xmath117 where @xmath118 and @xmath119 denotes the multiplicity of the eigenvalue @xmath120 .
the zeta function is holomorphic for @xmath121 and admits a meromorphic extension to @xmath122 with @xmath123 being a regular point .
consequently , the _ graded zeta - function _
@xmath124 is regular at @xmath123 and we may introduce the following [ graded - determinant ] let @xmath66 be an agmon angle for @xmath83 . then the graded determinant associated to @xmath83 and its agmon angle @xmath125
is defined as follows : @xmath126 [ rho - element ] @xcite .
the element @xmath127}\in \textup{det}(h^*(\domm , \dd))\ ] ] is independent of the choice of @xmath67 and choice of agmon angle @xmath66 for the odd - signature operator @xmath83 .
the construction of @xmath128 is in fact independent of the choice of a hermitian metric @xmath129 .
indeed , a variation of @xmath129 does not change the odd - signature operator @xmath65 as a differential operator and different hermitian metrics give rise to equivalent @xmath15norms over compact manifolds .
hence @xmath130 is indeed independent of the particular choice of @xmath129 .
independence of the choice of a hermitian metric @xmath129 is essential , since for non - unitary flat vector bundles there is no canonical choice of @xmath129 and a hermitian metric is fixed arbitrarily .
the refined analytic torsion is then obtained by studying the dependence of @xmath128 on the riemannian metric .
we cite the final result from @xcite .
[ rat - sign ] let @xmath131 be an odd - dimensional oriented compact riemannian manifold with boundary .
let @xmath132 be a flat complex vector bundle over @xmath6 .
consider the trivial vector bundle @xmath133 with a trivial connection @xmath134 and let @xmath135 denote the associated odd - signature operator .
@xmath136 denotes the eta invariant of the even part @xmath137 .
put @xmath138 then the refined analytic torsion of @xmath139 @xmath140\end{aligned}\ ] ] is modulo sign independent of the choice of @xmath47 in the interior of @xmath6 .
in the next step we interpret @xmath141 as an analytic section of the determinant line bundle over the representation space .
this requires a separate discussion of the analyticity notion for the refined torsion element and the graded determinant .
the present section studies analyticity of the refined torsion element , while the next deals with analyticity of the graded determinant .
the space @xmath142 of complex @xmath143-dimensional representations of @xmath144 has a natural structure of a complex analytic space , cf .
, for example , @xcite . for each @xmath145
we denote by @xmath146 the flat vector bundle over @xmath6 whose monodromy is equal to @xmath21
. then the disjoint union @xmath147 has a natural structure of a holomorphic line bundle over @xmath148 , called the _ determinant line bundle_. in this section we describe this structure , using a cw - decomposition of @xmath6 .
then , we show that the refined analytic torsion is a nowhere vanishing holomorphic section of @xmath149 .
we continue in the notation fixed in [ explicit - unitary ] .
fix a cw - decomposition @xmath150 of @xmath6 .
let @xmath151 denote the universal cover of @xmath152 .
then the fundamental group @xmath153 acts on @xmath154 from the right and @xmath4 is a left module over the group ring @xmath155 $ ] via the representation @xmath21 .
then the cochain complex @xmath156 is defined as @xmath157 } \c^n.\end{aligned}\ ] ] for each cell @xmath158 , fix a lift @xmath159 , a cell of the cw - decomposition of @xmath160 , such that @xmath161 . by definition ,
the pull - back of the bundle @xmath146 to @xmath160 is the trivial bundle @xmath162 .
hence , the choice of the cells @xmath163 identifies the cochain complex @xmath156 of the cw - complex @xmath152 with coefficients in @xmath164 with the complex @xmath165 where @xmath166 ( @xmath167 ) is equal to the number of @xmath168-dimensional cells of @xmath152 and the differentials @xmath169 are @xmath170-matrices depending analytically on @xmath145 .
the cohomology of the complex is canonically isomorphic to @xmath171 .
let @xmath172 denote the canonical isomorphism , cf .
formula ( 2.13 ) of @xcite .
the standard bases of @xmath174 ( @xmath175 ) define an element @xmath176 , and , hence , an isomorphism @xmath177 with @xmath178 .
then for each @xmath145 we define @xmath179 a non - zero element of @xmath180 .
of course , this element depends on the choice of the lifts @xmath163 .
let now @xmath182 denote the cw - decomposition of @xmath183 induced by @xmath152 .
then @xmath184 and the choice of the lifts @xmath185 made above identify the cochain complex @xmath186 of the cw - complex @xmath182 with coefficients in @xmath164 with the complex @xmath187 as above , the standard bases of @xmath188 ( @xmath189 ) defines a canonical element of @xmath190 and an isomorphism @xmath191 from @xmath192 onto @xmath193 .
thus we define @xmath194 where @xmath195 is the restriction of the representation @xmath21 to @xmath196 and @xmath197 is the restriction of @xmath146 to @xmath183 . consider the quotient complex @xmath198 using and we can identify @xmath199 with @xmath192 thus constructing a map @xmath200 the cohomology of the complex @xmath201 is canonically isomorphic to the relative cohomology @xmath202 . for each @xmath21
we define a non - zero element of @xmath203 by the formula @xmath204 recall that the elements @xmath205 and @xmath206 are defined in and . consider the map @xmath207 where @xmath145 , is a nowhere vanishing section of the determinant line bundle @xmath149 over @xmath208 .
[ d : holsection ] we say that a section @xmath209 of @xmath149 is _ holomorphic _ if there exists a holomorphic function @xmath210 on @xmath208 , such that @xmath211 .
this defines a holomorphic structure on @xmath149 , which is independent of the choice of the lifts @xmath163 of @xmath212 , since for a different choice of lifts the section @xmath213 will be multiplied by a constant . in the next subsection we show that this holomorphic structure is also independent of the cw - decomposition @xmath152 of @xmath6 . the choice of the lifts @xmath163 of @xmath212 determines an _ euler structure _
@xmath214 on @xmath6 , while the ordering of the cells @xmath212 determines a cohomological orientation @xmath215 , cf .
moreover , every euler structure and every cohomological orientation can be obtained in this way .
the farber - turaev torsion @xmath216 , corresponding to the pair @xmath217 , is , by definition , @xcite , equal to the element @xmath205 defined in . since the farber - turaev torsion is independent of the choice of the cw - decomposition of @xmath6 , cf .
@xcite , we conclude that the element @xmath205 is also independent of the cw - decomposition , but only depends on the euler structure and the cohomological orientation .
the lifts @xmath159 and the ordering of the cells also defines an euler structure @xmath218 and a cohomological orientation @xmath219 of @xmath183 .
the element @xmath220 defined in is equal to the farber - turaev torsion @xmath221 where @xmath195 is the restriction of the representation @xmath21 to @xmath196 .
let @xmath222 denote the fusion isomorphism , cf .
@xcite . using this isomorphism
we define the map @xmath223 it follows from the construction that the elements @xmath205 , @xmath220 and @xmath206 defined in , , , and satisfy the equality @xmath224 where the sign depends only on the dimensions of the spaces @xmath225 and @xmath226 but not on the representation @xmath21 . since @xmath205 and @xmath220 are independent of the cw - decomposition , it follows that @xmath206 is independent of cw - decompositon up to a sign .
in fact @xmath227 can be considered as the definition of the relative farber - turaev torsion .
it follows that the section @xmath213 defined in is also an independent of the cw - decomposition up to a sign .
hence so is the holomorphic structure defined in .
let @xmath228 denote the set of representations such that @xmath229 , @xmath230 , and @xmath231 .
then the determinant lines @xmath232 , @xmath233 , and @xmath234 are canonically isomorphic to @xmath192 .
hence , the farber - turaev torsions @xmath235 and @xmath236 can be viewed as a complex - valued functions on @xmath237 .
it is easy to see , cf .
* theorem 4.3 ) , that these functions are holomorphic on @xmath237 .
moreover , they are rational functions on @xmath208 , all whose poles are in @xmath238 in particular , the holomorphic structure on @xmath149 , which we defined above , coincides , when restricted to @xmath237 , with the natural holomorphic structure obtained from the canonical isomorphism @xmath239
fix @xmath240 . fix a number @xmath241 which is not in the spectrum of the square @xmath242 of the odd signature operator @xmath243 .
then there is a neighborhood @xmath244 of @xmath245 such that @xmath246 is not an eigenvalue of @xmath247 for all @xmath248 .
denote by @xmath249 the restriction of @xmath250 to the spectral subspace of @xmath247 corresponding to the spectral set @xmath251 .
then @xmath249 is an invertible operator .
let @xmath252 be an agmon angle for @xmath253 and that there are no eigenvalues of @xmath254 in the solid angles @xmath255}$ ] and @xmath256}$ ] .
then there exists an neighborhood @xmath257 of @xmath245 such that @xmath258 is also an agmon angle for @xmath259 for all @xmath260 . in this section
we prove that the graded determinant @xmath261 is a holomorphic function on @xmath262 .
our main result in this section is the following [ t : analytic ] let @xmath263 be a connected open neighborhood of @xmath264 .
let @xmath265 be a holomorphic curve such that @xmath266 .
then the function @xmath267 is holomorphic in a neighborhood of @xmath264 .
an immediate consequence is the following [ c : analytic ] suppose @xmath268 is an open subset such that all points @xmath269 are regular points of the complex algebraic set @xmath208
. then the map @xmath270 is holomorphic . by hartogs theorem ( cf .
, for example , ( * ? ? ?
2.2.8 ) ) , a function on a smooth algebraic variety is holomorphic if its restriction to each holomorphic curve is holomorphic .
hence , the corollary follows immediately from .
the rest of this section is occupied with the proof of theorem [ t : analytic ] .
let us introduce some additional notations .
let @xmath271 be a vector bundle over @xmath6 and let @xmath272 denote the affine space of ( not necessarily flat ) connections on @xmath271 .
we endow @xmath272 with the the frchet topology on @xmath272 introduced in section 13.1 of @xcite .
fix a base point @xmath273 and let @xmath274 denote the fiber of @xmath271 over @xmath275 .
we will identify @xmath274 with @xmath276 and @xmath277 with @xmath153 . for @xmath278 and a closed path @xmath279\to m$ ] with @xmath280
, we denote by @xmath281 the monodromy of @xmath282 along @xmath283 .
note that , if @xmath282 is flat then @xmath284 depends only on the class @xmath285 $ ] of @xmath283 in @xmath153 .
hence , if @xmath282 is flat , then the map @xmath286 defines an element of @xmath208 , called the _
monodromy representation _ of @xmath282 .
suppose now that @xmath263 is a connected open neighborhood of @xmath264 .
let @xmath287 be a holomorphic curve with @xmath266 .
the operator @xmath288 is constructed using a flat connection @xmath289 whose monodromy is equal to @xmath290 .
unfortunately , there is no a canonical choice of such connection . though the graded determinant @xmath291 is independent of this choice , to study the dependence of this determinant on @xmath292 we need to choose a family of connections @xmath289 .
the main difficulty in the proof of theorem [ t : analytic ] is that it is not clear whether there exists a _
family @xmath289 with @xmath293 .
we shall now explain how to circumvent this difficulty . by proposition 4.5 of @xcite ,
all the bundles @xmath294 , are isomorphic to each other . in other words , there exists a vector bundle @xmath295 and a family of flat connections @xmath296 , on @xmath271 , such that the monodromy representation of @xmath289 is equal to @xmath290 for all @xmath292 .
moreover , ( * ? ? ?
* lemma b.6 ) shows that the family @xmath289 can be chosen so that there exists a one - form @xmath297 such that @xmath298 where @xmath299 is understood in the sense of the frchet topology on @xmath272 introduced in section 13.1 of @xcite , and always refers to the behaviour as @xmath300 .
let @xmath301 denote the set of flat connections on @xmath271 and consider a curve @xmath302 , @xmath292 , of connections such that @xmath303 .
we will assume that it is complex differentiable at 0 in the sense that @xmath304 where @xmath305 .
we denote by @xmath306 and by @xmath307 the corresponding odd signature operator the sense of [ explicit - unitary ] .
the space @xmath308 of @xmath14-differential forms is a direct sum @xmath309 consider the corresponding orthogonal projections @xmath310 let @xmath311 denote the spectral projection of the operator @xmath312 onto the subspace @xmath313 spanned by eigenforms of @xmath312 with eigenvalues in @xmath251 .
then @xmath314 and @xmath315 commute with @xmath316 .
hence , @xmath317 are projections and @xmath318 [ fam - proj ] there exist bounded operators @xmath319 on @xmath308 with @xmath320 for small enough @xmath292 the operators @xmath321 depend smoothly on @xmath322 , there exist bounded operators @xmath319 and @xmath323 such that @xmath324 using the decomposition @xmath325 and the equality @xmath326 we obtain @xmath327 comparing and and using we conclude that @xmath328 the equality follows now from and . in terms of the projections @xmath321 introduced in the previous section
the odd signature operator @xmath329 can be written in the form @xmath330 hence , we may write @xmath331 consider a curve @xmath332 such that @xmath333 . to simplify the notation , set @xmath334 and also @xmath335 with this notation we have @xmath336 we consider the function @xmath337 notice that the right hand side of is similar to the right hand side of but @xmath321 is replaced by @xmath338 . in particular , @xmath339 then for any curve @xmath332 with @xmath333 we have @xmath340 using lemma [ fam - proj ] and the equality @xmath341 we obtain @xmath342 by the variation formula for the logarithm of the determinant , cf . , for example , section 3.7 of @xcite , we have @xmath343 using and the fact that the operators @xmath344 and @xmath345 commute with @xmath346 we conclude that @xmath347 combining and we conclude that @xmath348 similarly , @xmath349 from , , and we obtain .
[ p : analyticn ] let @xmath350 @xmath351 be a family of flat connections such that as @xmath300 @xmath352 then the function @xmath353 is complex differentiable at zero .
in other words , there exists a complex number @xmath323 such that @xmath354 by it is enough to show that @xmath355 set @xmath356 . using the variation formula for the logarithm of the determinant as in the proof of lemma [ l : partialdet ]
, one easily sees that @xmath357 which is equivalent to .
from @xmath358 it follows from proposition [ p : analyticn ] that the function @xmath359 is complex differentiable at 0 .
let @xmath360 be such that @xmath361 . by making a change of variables
@xmath362 we conclude that this function is also complex differentiable at @xmath363 .
hence , this function is holomorphic in @xmath364 . @xmath365
in this section we show that the refined analytic torsion @xmath366 is a non - vanishing holomorphic section of @xmath149 .
more precisely , our main result in the following [ t : holomsection ] the refined analytic torsion @xmath366 is a holomorphic section of the determinant bundle @xmath149 . in particular , the restriction of @xmath366 to the set @xmath237 of acyclic representations , viewed as a complex - valued function via the canonical isomorphism @xmath367 is a holomorphic function on @xmath237 .
let @xmath245 be a regular point in @xmath208 .
fix a riemannian metric @xmath47 on @xmath6 and a number @xmath241 such that there are no eigenvalues of @xmath368 with absolute value equal to @xmath246 .
let @xmath258 be an agmon angle for @xmath368 and let @xmath369 be as in .
by corollary [ c : analytic ] the function @xmath370 is holomorphic on @xmath262 .
it follows now from the definition of the refined analytic torsion that to prove it is enough to show that @xmath371}\ \equiv\ \rho_{\gg_{[0,\lam]}}(\na , g)\ ] ] is a holomorphic section of @xmath149 . by the definition of the holomorphic structure on the bundle @xmath149 , cf .
, this section is holomorphic if and only if the function @xmath372}}(\na , g)}{\tau(\alp)},\ ] ] is holomorphic . here
, @xmath213 is defined in . by hartog
s theorem , ( * ? ? ?
2.2.8 ) , it is enough to show that for every holomorphic curve @xmath373 , where @xmath374 is a connected open neighborhood of 0 in @xmath192 , the function @xmath375}}(\n_{\gam(z)},g)}{\tau(\gam(z))}\ ] ] is complex differentiable at @xmath264 , i.e. , there exists @xmath376 , such that as @xmath300 @xmath377 we use the notation introduced in .
in particular we have a vector bundle @xmath271 and a family of flat connection @xmath378 on @xmath271 such that for each @xmath292 the monodromy of @xmath289 is equal to @xmath290 and @xmath379 where @xmath299 is understood in the sense of the frchet topology .
let @xmath380}(z)$ ] ( @xmath292 ) denote the spectral projection of the operator @xmath381 , corresponding to the set of eigenvalues of @xmath381 , whose absolute value is @xmath382 .
then @xmath380}(z)$ ] is a complex differentiable family of bounded operators .
hence , there exists a bounded operator @xmath383 on @xmath75 such that @xmath384}(z ) \ = \
\pi_{[0,\lam]}(0 ) \ + \ z\ ,
o(z).\ ] ] we denote by @xmath385 the image of @xmath380}(z)$ ] .
recall that we denote the dimension of @xmath6 by @xmath386 . for each @xmath387 ,
fix a basis @xmath388 of @xmath389 and set @xmath390 . to simplify the notation we will write @xmath391
. then @xmath392 is a basis for @xmath389 for all @xmath175
. for each @xmath292 , @xmath175 , set @xmath393}(z)\,w^1_j\nek \pi_{[0,\lam]}(z)\,w^{l_j}_j\,\big\}.\ ] ] it follows from the definition of @xmath394 that the projection @xmath380}(z)$ ] depends continuously on @xmath322 .
hence , there exists a neighborhood @xmath395 of @xmath264 , such that @xmath396 is a basis of @xmath397 for all @xmath398 , @xmath175 .
further , since @xmath380}(z)$ ] commutes with @xmath60 , we obtain @xmath399 clearly , @xmath400 for all @xmath175 .
let @xmath401 denote the canonical isomorphism , cf .
section 2.4 of @xcite . for @xmath398 , let @xmath402 be the element determined by the basis @xmath403 of @xmath385 . more precisely , we introduce @xmath404 and set @xmath405 then , according to definition 4.2 of @xcite , it follows from that , for all @xmath398 , the refined torsion of the complex @xmath385 is equal to @xmath406 , i.e. , @xmath407}}}}(\n_{\gam(z ) } ) \ = \
\phi_{\ome^\b(z)}(w(z)).\ ] ] for each @xmath398 , the space @xmath385 is a subcomplex of @xmath408 , whose cohomology is canonically isomorphic to @xmath409 using the basis @xmath396 we define the isomorphism @xmath410}(z)\ ] ] by the formula @xmath411}(z)\,w_j^k.\ ] ] we conclude that for each @xmath398 , the complex @xmath412 is isomorphic to the complex @xmath413 where @xmath414 it follows from and that @xmath415 is complex differentiable at @xmath264 , i.e. , there exists a @xmath416-matrix @xmath323 such that @xmath417 let @xmath418 .
since @xmath419 ( @xmath175 ) , we conclude that @xmath420 . from
we obtain that @xmath421 is independent of @xmath398 and @xmath422 it follows from and that @xmath423}}}}(\n_{\gam(z)}),\ ] ] where @xmath424 denotes the refined torsion of the finite dimensional complex @xmath425 corresponding to the chirality operator @xmath426 .
let @xmath427 denote the denote the canonical isomorphism of section 2.4 of @xcite .
the standard bases of @xmath428 ( @xmath175 ) define an element @xmath429 . from and
the definition of @xmath430 we conclude that @xmath431 to illustrate the main idea of the proof let us first consider the case , when both @xmath432 and @xmath433 are trivial .
then there exists a neighborhood @xmath434 of @xmath264 such that @xmath435 for all @xmath436 .
thus the torsion is a complex valued function on @xmath437 .
to finish the proof of theorem [ t : holomsection ] in this case it remains to show that this function is complex differentiable at @xmath264 . in view of , this follows from the following [ l : asyclicanalytic ] let @xmath438 be a family of acyclic complexes defined for all @xmath322 in an open set @xmath263 .
suppose that the differentials @xmath439 are complex differentiable at @xmath264 .
then for any @xmath440 the function @xmath441 is complex differentiable at @xmath264 .
it is enough to prove the lemma for one particular choice of @xmath442 . to make such a choice
let us fix for each @xmath175 a complement of @xmath443 in @xmath444 and a basis @xmath445 of this complement .
since the complex @xmath446 is acyclic , for all @xmath175 , the vectors @xmath447 form a basis of @xmath444 .
let @xmath440 be the element defined by these bases .
then , for all @xmath322 close enough to @xmath264 and for all @xmath175 , @xmath448 is also a basis of @xmath444 .
let @xmath449 denote the non - degenerate matrix transforming the basis to the basis .
then , by the definition of the isomorphism @xmath450 , cf .
2.4 of @xcite , @xmath451 where @xmath452 is the integer defined in formula ( 2.15 ) of @xcite which is independent of @xmath322
. clearly , the matrix valued functions @xmath453 and , hence , their determinants are complex differentiable at @xmath264 .
thus , so is the function @xmath454 .
we now turn to the proof of theorem [ t : holomsection ] in the general case . in this subsection
we sketch the main ideas of the proof .
it is enough to show that the function @xmath455}}}}(\n_{\gam(z)},g)}{\tau(\gam(z))}\ ] ] is complex differentiable at @xmath264 , where @xmath456 is the map . to see this we consider the de rham integration maps @xmath457 where the cochain complexes @xmath458 and @xmath459 are defined in [ s : detlinebundle ] . the de rham integration map of @xmath271-valued differential forms is defined using a trivialization of @xmath271 over each cell @xmath158 , and , hence , it depends on the flat connection @xmath289 , cf .
more precisely , in the neighborhood of any cell @xmath460 of @xmath152 , a differential form @xmath461 can be written in the form @xmath462 , where @xmath463 is a complex - valued differential form and @xmath464 is a @xmath465-flat section of @xmath466 for @xmath467 .
the de rham integration map is then defined by @xmath468 if @xmath469
then @xmath470 is a well - defined element of the relative cochain complex @xmath459 and we denote the corresponding map by @xmath471 . both maps descend to isomorphisms on cohomology @xcite .
we write @xmath472 hence we obtain a quasi - isomorphism @xmath473 the trivialization @xmath474 induced by the connection @xmath465 is complex differentiable at @xmath475 by , and hence so is the de rham integration map @xmath476 .
we then consider the cone complex @xmath477 of the map @xmath476 .
this is a finite dimensional acyclic complex with a fixed basis , obtained from the bases of @xmath385 , defined [ ss : basis ] , and @xmath478 .
the torsion of this complex is equal to @xmath479 .
an application of to this complex proves theorem [ t : holomsection ] . in the definition of the integration map @xmath476 we have to take into account the fact that the vector bundles @xmath466 and @xmath480 are isomorphic but not equal . the integration map @xmath476 , cf .
, is a map from @xmath481 to the cochain complex @xmath478 , which is not equal to the complex @xmath482 .
fix an euler structure on @xmath6 .
it defines an isomorphism between the complexes @xmath482 and @xmath478 which depends on @xmath322 .
the study of this isomorphism , which is conducted in , is important for the understanding of the properties of @xmath476 . in particular
, it is used to show that @xmath476 is complex differentiable at @xmath264 , which implies that the cone complex @xmath477 satisfies the conditions of .
fix a cw - decomposition @xmath483 of @xmath6 . for
each @xmath484 choose a point @xmath485 and let @xmath486 denote the fiber of @xmath271 over @xmath487 .
then the cochain complex @xmath488 may be naturally identified with the complex @xmath489 where the @xmath322-dependence is now fully encoded in the differentials .
we use the prime in the notation of the differentials @xmath490 in order to distinguish them from the differentials of the cochain complex @xmath478 .
it follows from that @xmath491 are complex differentiable at @xmath264 , i.e. , there exist linear maps @xmath492 , s.t .
@xmath493 recall that for each @xmath398 the monodromy representation of @xmath289 is equal to @xmath290 .
let @xmath494 denote the universal cover of @xmath6 and let @xmath495 denote the pull - back of the bundle @xmath271 to @xmath160 .
recall that in we fixed a point @xmath273 .
let @xmath496 be a lift of @xmath275 to @xmath160 and fix a basis of the fiber @xmath497 of @xmath498 over @xmath275 .
then , for each @xmath398 , the flat connection @xmath289 identifies @xmath498 with the product @xmath499
. recall from that the choice of the euler structure @xmath214 also fixes the lifts @xmath500 of the cell @xmath158 fixed in .
let @xmath501 be the lift of @xmath485 .
then the trivialization of @xmath498 defines isomorphisms @xmath502 the isomorphisms @xmath503 depend on the trivialization of @xmath498 , i.e. , on the connection @xmath289 .
the direct sum @xmath504 defines an isomorphism @xmath505 .
it follows from that @xmath506 is complex differentiable at @xmath264 , i.e. for some linear map @xmath507 @xmath508 finally , we consider the morphism of complexes @xmath509 this map is complex differentiable at @xmath475 and induces an isomorphism of cohomology .
the cone complex @xmath510 of the map @xmath511 is given by the sequence of vector spaces @xmath512 @xmath175 , with differentials @xmath513 where @xmath514 denotes the restriction of @xmath511 to @xmath515 .
this is a family of acyclic complexes with differentials @xmath516 , which are complex differentiable at @xmath264 .
the standard bases of @xmath517 define an element @xmath518 which is independent of @xmath398 . using the canonical isomorphism of section 2.4 of @xcite ,
we hence obtain for each @xmath398 the number @xmath519 . from the discussion in it
follows that this number is equal to the ratio . hence , to finish the proof of the theorem [ t : holomsection ] it remains to show that the function @xmath520 is complex differentiable at @xmath264 .
this follows immediately from .
let @xmath131 be a closed oriented riemannian manifold and @xmath521 a separating hypersurface , such that @xmath522 .
the metric @xmath47 restricts to riemannian metrics on the two compact components @xmath523 and @xmath524 .
assume that @xmath47 is product in an open tubular neighborhood of @xmath525 .
a given representation @xmath527 induces a connection @xmath528 on a vector bundle @xmath529 , which restricts to well - defined connections on @xmath530 .
we denote by @xmath531 the refined analytic torsions on @xmath532 and by @xmath533 for the refined analytic torsion on @xmath6 , cf . .
let @xmath534 denote the obvious inclusions , @xmath535 .
we define the transmission subcomplex of @xmath536 by specifying _ transmission _ boundary conditions @xmath537 this defines a complex with eigenforms of the corresponding laplacian given by the eigenforms of the hodge - laplacian on @xmath538 , cf .
* theorem 5.2 ) .
in particular their de rham cohomologies coincide .
the splitting @xmath539 now gives rise to short exact sequences of the associated complexes @xmath540 where @xmath541 and @xmath542 .
the corresponding long exact sequences in cohomology yield canonical isomorphisms , cf .
@xcite @xmath543 the fusion isomorphisms , cf .
* ( 2.18 ) ) provide canonical identifications @xmath544 where the hilbert complexes @xmath545 are defined in definition [ domain ] , with the lower index @xmath168 referring to the underlying manifold @xmath546 .
the hilbert complex @xmath63 is defined over @xmath6 .
we put @xmath547 if @xmath548 is unitary , ( * ? ? ?
* theorem 10.6 ) asserts @xmath549 where @xmath550 here @xmath551 stands for the euler characteristic of @xmath552 and the @xmath553-invariants @xmath554 , @xmath555 and @xmath556 are defined in terms of the even parts of the corresponding odd signature operators .
the sign @xmath557 is determined by formula ( 8.4 ) of @xcite .
note that the sign depends on the dimensions of various cohomology spaces and is related to the sign convention used in defining the fusion isomorphism of determinant lines , see @xcite for a detail discussion of the sign conventions .
note also that the euler characteristic @xmath551 depends only on the rank @xmath143 of the representation @xmath21 but not on the particular choice of @xmath145 .
the @xmath553-invariants @xmath554 , @xmath555 and @xmath556 are not necessarily continuous functions of @xmath21 : they have jumps when some eigenvalues of the odd signature operator cross zero .
however , we will show below that @xmath558 extends to a holomorphic function on the space @xmath208 of representations .
the main result of this section is the following extension of to some class of non - unitary representations .
[ refined - gluing - theorem ] let @xmath131 be a closed oriented riemannian manifold of odd dimension , and @xmath525 a separating hypersurface such that @xmath522 , and @xmath47 is product in an open tubular neighborhood of @xmath525 .
assume that @xmath559 is a connected component and @xmath560 is a unitary representation which is a regular point of the complex analytic set @xmath561 .
for any representation @xmath562 denote by @xmath141 and @xmath563 the refined analytic torsions on @xmath6 and @xmath546 .
then for any @xmath564 we have @xmath565 the rest of this section is occupied with the proof of theorem [ refined - gluing - theorem ] , which is based on an analytic continuation technique , cf .
first we need the following proposition .
[ k ] @xmath566 is a holomorphic function on the complex analytic space @xmath208 . by ( * ? ?
* ( 4.6 ) ) we obtain the following formula for the graded determinant on @xmath567 @xmath568 where we have set @xmath569 analyticity of the graded determinant is established in proposition [ p : analyticn ] .
analyticity for @xmath570 and @xmath571 follows similarly from the variational formula @xmath572 as a consequence , @xmath573 and @xmath574 must also be holomorphic .
the set of unitary representations is the fixed point set of the anti - holomorphic involution @xmath575 where @xmath195 denotes the representation dual to @xmath21 .
hence , it is a totally real submanifold of @xmath208 whose real dimension is equal to @xmath576 , cf .
* proposition 3 ) .
in particular there is a holomorphic coordinates system @xmath577 near @xmath245 such that the unitary representations form a _ real neighborhood _ of @xmath245 , i.e. the set @xmath578 .
therefore , cf .
@xcite , we obtain the following [ p : continuation ] if two holomorphic functions coincide on the set of unitary representations they also coincide on @xmath561 . by theorem [ t : holomsection ]
the refined analytic torsions @xmath579 and @xmath580 define holomorphic sections on the corresponding determinant line bundles .
the canonical isomorphism @xmath581 defines a bilinear map between holomorphic determinant line bundles and hence maps holomorphic sections to holomorphic sections .
let us denote by @xmath210 the unique complex valued function of @xmath21 such that @xmath582 holds for all @xmath564 .
since both @xmath583 and @xmath580 are holomorphic sections of the determinant line bundle , @xmath210 is a holomorphic function .
it follows from that @xmath584 for all unitary representations in @xmath561 . by proposition [ k ]
, @xmath566 is a holomorphic function .
hence , we obtain from proposition [ p : continuation ] that holds for all @xmath564 .
@xmath365 [ r : ref - usual ] the gluing formula for refined analytic torsion may be used to prove a gluing result for the ray - singer torsion norm on connected components of the representation variety that contain a unitary point .
we continue in the previously outlined setup of a closed oriented riemannian manifold @xmath131 and a separating hypersurface @xmath521 , such that @xmath539 . consider a representation @xmath145 and the corresponding flat vector bundle @xmath585 .
fix a hermitian metric @xmath586 on @xmath271 , of product type near @xmath525 . even if the metric structures @xmath587 are product near @xmath525 , the connection @xmath588 need not be product near @xmath525 , so that the resulting laplacian does not have a product structure near the separating hypersurface and hence a gluing theorem for ray - singer analytic torsion can not be obtained from the results of @xcite , @xcite and @xcite .
before we proceed , let us make some chronological remarks on that topic .
vishik @xcite was first to prove the gluing formula for analytic torsion given a unitary representation without using the theorem of cheeger @xcite and mller @xcite .
though it was not explicitly stated in @xcite , the assumption of a unitary representation is obsolete once a connection is in temporal gauge near @xmath525 . under the assumption of temporal gauge and product metric structures ,
the hodge laplacian is of product type near @xmath525 and the vishik s argument goes through .
recently , lesch @xcite provided an excellent discussion of the gluing formula for possibly non - compact spaces , extending the result of vishik , and stating clearly that the proof requires only product metric structures and a connection in temporal gauge near the cut , rather than unitariness of the representation .
a quite general proof of the gluing formula for general representations and without assuming product metric structures , was provided by brning and ma @xcite .
they derive a gluing formula by relating the ray - singer and the milnor torsions , in odd and also in the even - dimensional case . in this section
we present a different proof of their result in odd dimensions , as a consequence of @xcite and the brning - ma anomaly formula in @xcite . to make the main idea of our alternative argument clear , we restrict ourselves to the situation , when @xmath589 .
[ bm - hidden ] consider two hermitian metrics @xmath590 on a fixed flat vector bundle @xmath585 over a compact oriented odd - dimensional riemannian manifold @xmath591 . if @xmath592 , assume that @xmath593 coincide over @xmath594 .
fix either relative or absolute boundary conditions at @xmath595 for the hodge laplacian and denote by @xmath596 the corresponding ray - singer analytic torsion norms . then and @xmath597 . ]
@xmath598 the metric anomaly , identified in @xcite is expressed in terms of the levi - civita connections @xmath599 and @xmath600 on @xmath152 and its boundary @xmath595 , the respective representatives @xmath601 of the euler classes of @xmath602 in chern - weil theory , and the quotient @xmath603 between the metrics on @xmath604 , induced by @xmath586 and @xmath605 . since @xmath606 is odd , @xmath607 . if @xmath593 coincide over @xmath595 , @xmath608 over @xmath595 , so that the statement follows from ( * ? ? ?
* theorem 0.1 , ( 0.5 ) ) .
the main idea now is the reduction to the setup of a connection in temporal gauge near @xmath525 .
a connection @xmath588 is in temporal gauge in an open neighborhood @xmath609 of @xmath525 , if @xmath610 for some flat connection @xmath611 on @xmath612 , where @xmath613 is the natural projection onto the second factor .
proposition [ t - g ] below asserts that in fact every connection is gauge equivalent to a connection in temporal gauge .
we denote the corresponding gauge transformation by @xmath614 . the gauge transformed connection is given by @xmath615 .
we set for any @xmath616 @xmath617 this defines a new hermitian metric on @xmath271 that coincides with @xmath618 over @xmath525 , since @xmath614 acts as identity over @xmath525 . by construction
, @xmath614 induces an isometry @xmath619 the following theorem is a result by lesch @xcite , cf .
vishik @xcite .
[ lesch ] let @xmath131 be a closed oriented riemannian manifold of odd dimension , and @xmath525 a separating hypersurface such that @xmath522 , and @xmath47 is product in an open tubular neighborhood of @xmath525 .
consider flat hermitian vector bundle @xmath620 .
denote by @xmath621 and @xmath622 the corresponding ray - singer norms on @xmath6 and @xmath623 , respectively .
the ray - singer norm on @xmath523 is defined with respect to relative boundary conditions , while on @xmath524 we pose absolute boundary conditions .
then @xmath624 let the ray - singer torsion norms @xmath621 and @xmath622 be defined with respect to the flat connection @xmath588 and any hermitian metric @xmath618 on @xmath271 .
then @xmath624 the laplacian @xmath625 on @xmath626 is related to the laplacian @xmath627 on @xmath628 by the unitary transformation @xmath614 with @xmath629 .
hence , @xmath614 induces a map between the harmonic forms of @xmath627 and @xmath625 , and hence also between the corresponding determinant lines , which we also denote by @xmath614 . by construction
, we find @xmath630 where we indicate the dependence on the vector bundle by the subindex and omit the reference to the underlying manifold , since the relation holds both on @xmath6 and @xmath532 .
the isometric identification @xmath614 commutes with the maps in the .
hence @xmath631 , and by theorem [ lesch ] , we find @xmath632 this is a gluing theorem for any complex representation @xmath21 , possibly non - unitary . a priori , however , this relation holds for the specific hermitian metric @xmath633 on @xmath634 . since @xmath635 ,
the metrics @xmath636 coincide over @xmath525 and hence , by proposition [ bm - hidden ] the gluing formula holds for any representation @xmath21 and any choice of a hermitian metric @xmath618 on @xmath634 . the riemannian metric @xmath47 is still assumed to be product near @xmath525 , so that variation of @xmath47 leads to additional anomaly terms , cf .
* theorem 0.1 ) .
in this section we recall the notion of a connection in temporal gauge and review some properties of these connections , cf . @xcite .
in particular , we show that any flat connection is gauge equivalent to a connection in temporal gauge , cf .
proposition [ t - g ] .
consider a closed oriented riemannian manifold @xmath637 of dimension @xmath638 and a vector bundle @xmath271 with structure group @xmath639 .
denote the principal @xmath640-bundle associated to @xmath271 by @xmath24 , where @xmath640 acts on @xmath24 from the right .
consider a hypersurface @xmath641 and its collar neighborhood @xmath642 .
we view the restrictions @xmath643 as @xmath640-bundles , where the structure group can possibly be reduced to a subgroup of @xmath640 .
let @xmath644 be the natural projection onto the second component .
then @xmath645 and for the associated principal bundles @xmath646 , where @xmath647 is the principal bundle homomorphism , covering @xmath648 , with the associated homomorphism of the structure groups being the identity automorphism .
we call a flat connection @xmath116 on @xmath24 a connection in temporal gauge over @xmath649 , if there exists a flat connection @xmath650 on @xmath651 such that @xmath652 over the collar neighborhood @xmath649
. similar condition defines a covariant derivative in temporal gauge .
we now explain the notion of temporal gauge in local terms .
let @xmath653 denote a flat connection one - form on @xmath654 .
then @xmath655 gives a connection one - form on @xmath656 which is flat again .
let @xmath657 be a set of local trivializations for @xmath654
. then @xmath658 is trivialized over the local neighborhoods @xmath659 with the induced trivializations @xmath660 . for any @xmath661 , normal variable @xmath662 and for @xmath663 being the identity matrix we put @xmath664
these local sections define local representations for @xmath665 and @xmath653 @xmath666 where @xmath667 denotes the lie algebra of @xmath640 .
consider local coordinates @xmath668 on @xmath669 .
then @xmath670 by a partition of unity argument it suffices to discuss the problem locally over @xmath671 .
let @xmath116 be a flat connection on @xmath656 .
let @xmath672 be any gauge transformation on @xmath656 and @xmath673 the corresponding local representation .
denote the gauge transform of @xmath116 under @xmath614 by @xmath674 .
the local @xmath667-valued one - forms @xmath675 are related in correspondence to the transformation law of connections by @xmath676 where the action @xmath677 is the multiplication of matrices ( @xmath639 ) , after evaluation at a local vector field and a base point in @xmath671 .
the local one form @xmath678 can be written as @xmath679 our task is to identify the correct gauge transformation @xmath614 , so that @xmath116 is temporal gauge , cf . .
for this reason we consider the following initial value problem with parameter @xmath680 @xmath681 the solution to is given by an integral curve of the time dependent vector field @xmath682 on @xmath640 , parametrized by @xmath683 , such that for any @xmath684 @xmath685 where @xmath686 is the right multiplication on @xmath614 and the second equality follows from the fact that @xmath639 is a matrix lie group .
the corresponding integral curve @xmath687 with @xmath688 is @xmath640-valued and the unique solution to .
we now compute for the gauge transformed connection @xmath674 @xmath689 where in the last equality we cancelled two summands due to @xmath673 being the solution to .
we now use the fact that @xmath116 is a flat connection .
a gauge transformation preserves flatness , so @xmath674 is flat again . put @xmath690 where by the previous calculation @xmath691 flatness of @xmath674 implies @xmath692
hence the gauge transformed connection is indeed in temporal gauge .
this completes the proof .
shabat , b. v. , _ introduction to complex analysis .
part ii _ , translations of mathematical monographs , vol .
110 , american mathematical society , providence , ri , 1992 , functions of several variables , translated from the third ( 1985 ) russian edition by j. s. joel . | we prove that refined analytic torsion on a manifold with boundary is an analytic section of the determinant line bundle over the representation variety . as a fundamental application
we establish a gluing formula for refined analytic torsion on connected components of the complex representation space which contain a unitary point .
finally we provide a new proof of brning - ma gluing formula for the ray - singer torsion associated to a non - hermitian connection .
our proof is quite different from the one given by brning and ma and uses a temporal gauge transformation . |
the derivation of fourier s law , or any other macroscopic law , from the microscopic underlying motion of particles constitutes a major task in condensed matter physics .
this task comprises not only the derivation itself but also the problem of setting up the appropriate microscopic model . the simplest model one could conceive to derive fourier s law is a chain of particles interacting through harmonic potentials in contact with two heat reservoirs at each end .
however , it has been shown by rieder et al @xcite that this model does not lead to fourier s law . since then , several microscopic models have been introduced and studied @xcite some of them leading instead to the so called anomalous fourier s law .
fourier s law states that the heat flux @xmath0 is proportional to the gradient of the temperature @xmath1 , that is , @xmath2 where @xmath3 is the heat conductivity . if we consider a small bar of length @xmath4 subject to a difference in temperature @xmath5 , then @xmath6 .
thus a microscopic model for fourier s law should predict a heat flux that decreases with @xmath4 , for a fixed value of @xmath5 , according to @xmath7 this amounts to saying that @xmath3 is finite .
if , instead , we find that @xmath8 with @xmath9 , then we are faced with the anomalous fourier s law . in this case
we may say that @xmath3 is infinite , as in the harmonic chain @xcite , diverging according to @xmath10 , with exponent @xmath11 .
notice that @xmath4 should be macroscopically small , so that equation ( [ 5 ] ) is the expression of fourier s law , but microscopically large so that a microscopic model for this law should yield equation ( [ 5 ] ) for sufficiently large @xmath4 .
the heat flux @xmath0 of the linear harmonic chain placed between two heat reservoirs has been shown @xcite to be independent of the size @xmath4 of the chain , meaning that the heat conductivity @xmath3 is infinite .
this result is a direct consequence of the ballistic transmission of heat by the elastic waves , from one reservoir to the other .
a consequence of this result is that a perfectly harmonic crystal has an infinite heat conductivity @xcite . in real crystals
the heat conductivity is manifestly finite .
this is due to the presence of lattice imperfections , impurities , and other factors that act as scattering centers for the waves carrying energy , such as the umklapp process @xcite .
these factors make the heat conduction a diffusion process which implies fourier s law .
the crossover from ballistic do diffusive behavior can also be undertood in terms of the dephasing of the elastic waves caused by the scattering events just mentioned .
this process is characterized by a dephasing length @xcite which , when smaller than the size of the system , gives the linear temperature profile expected from fourier s law .
thus , a possible ingredient in the microscopic derivation of fourier s law consists in the presence of a diffusive motion at the microscopic level .
one way of introducing this ingredient is by means of stochastic collisions that change the sign of the velocity of the particles .
this can be accomplished in the form of impulsive shot noises with exponentially distributed waiting times @xcite .
two key properties are required when devising such a noise .
firstly , it should conserve the total energy of the system because any variation of the energy of the system should only be due to the contact with the heat reservoirs . changing the sign of the velocity
does not alter the energy .
secondly , it should make the system ergodic even when it is not coupled to the heat baths .
a harmonic chain with this type of shot noise has been indeed studied by dhar et al @xcite .
they showed that this model can be mapped into the self consistent harmonic crystal introduced by bolsteri et al @xcite , and studied by bonetto et al @xcite . in this model
each particle is in contact with independent heat reservoirs , whose temperatures are chosen so that , in the steady state , there is no exchange of heat between these reservoirs and the chain .
the contact with the reservoirs is regarded as a procedure to make the system ergodic @xcite .
this model predicts a linear profile for the temperature and a finite heat conductivity , which is independent of temperature .
here we study a chain of particles interacting through anharmonic forces in addition to random reversals of the velocity .
more specifically we consider a potential with quartic terms in addition to the harmonic quadratic terms . without the stochastic shot noise
, this is the well known fermi - pasta - ulam model in contact with heat reservoirs at its ends , which was studied numerically by lepri et al .
@xcite who found a superdiffusive transport of heat implying an anomalous fourier s law
. this important result has been confirmed by other numerical studies and other approaches @xcite the impulsive stochastic shot noise that we use here can be regarded as a procedure that turns the superdiffusive transport of heat into a diffuse transport leading thus to fourier s law . by numerically solving the langevin equations for chains of several sizes , we determine the heat conductivity as a function of the system size @xmath4 and the rate of stochastic collisions @xmath12 .
when @xmath12 is nonzero , we obtain a finite heat conductivity and therefore fourier s law . for small values of @xmath12 ,
the heat conductivity is found to behave as as @xmath13 , diverging when @xmath14 .
our numerical results give @xmath15 .
we have also determined the exponent @xmath16 related to the divergence of @xmath3 with @xmath4 at @xmath14 and found @xmath17 .
these results are distinct from the harmonic case @xcite for which @xmath18 and @xmath19 .
also , in contrast to the harmonic case , we have found that the heat conductivity depends on temperature .
more precisely , for a fixed @xmath5 , we found that it increases monotonically with the temperatures of the reservoirs .
we consider a chain of @xmath4 interacting particles with equal masses @xmath20 and denote by @xmath21 the position of the @xmath22-th particle . the total potential energy @xmath23 )
is considered to be a sum of anharmonic potential energies involving nearest - neighbor pairs of particles @xmath24,\ ] ] where @xmath25 and @xmath26 are parameters .
we consider fixed boundary conditions so that @xmath27 .
when @xmath28 , the harmonic potential is recast .
the force @xmath29 acting on the @xmath22-th particle due to the potential @xmath30 is @xmath31 @xmath32 the first and last particles are connected to heat baths at temperatures @xmath33 and @xmath34 , and all particles are susceptive to stochastic collisions , here described by forces @xmath35 . denoting by @xmath36 the velocity of the @xmath22-th particle , the equations of motion are stochastic equations given by @xmath37 @xmath38 @xmath39 where @xmath40 is boltzmann constant and the last two terms in equations ( [ 1 ] ) and ( [ 3 ] ) represent the coupling to the heat reservoirs .
the constant @xmath41 represents the strength of the coupling and @xmath42 and @xmath43 are independent gaussian white noises with zero mean and unit variance .
the forces @xmath35 have the form of impulsive shot noises , given by @xmath44 where @xmath45 are uncorrelated exponentially distributed stochastic waiting times with a probability density distribution @xmath46 . here
, the parameter @xmath12 is the rate of collisions , which has been taken to be the same for all particles . after a collision occurring at time @xmath47
, the @xmath22-th particle changes its velocity from @xmath48 to @xmath49 , thus conserving its kinetic energy and therefore the total energy @xmath50 due to the contact with the reservoirs , the total energy @xmath51 is not a strictly conserved quantity . from the stochastic equations of motion we find @xmath52 where @xmath53 and @xmath54 in the stationary state , @xmath55 is a constant and the sum of the fluxes @xmath56 and @xmath57 vanishes , that is , @xmath58 .
the heat flux @xmath59 can be calculated as the average of the right - hand side of equation ( [ 13a ] ) or in a equivalent way from the equation @xmath60 which we found to be numerically more accurate than the formula @xmath61 .
the stochastic equations of motion were solved numerically for chains of several sizes @xmath4 .
this was done by discretizing the time in intervals @xmath62 .
we use an approach in which the deterministic part of the equations of motion of the inner particles are handled by the use of the verlet algorithm @xcite so as to ensure that , in the absence of the heat baths , energy is conserved . for the equations of motion of the first and last particles , which contain the stochastic forces due to the heat baths , we used the stochastic verlet algorithm developed in @xcite .
as for the stochastic shot noises we treat them as follows . at each time step
, the sign of the velocity of each particle is changed with probability @xmath63 .
this procedure generates a poisson process with discrete waiting time @xmath64 that is distributed according to the probability distribution @xmath65 . in the continuous time limit , @xmath66 ,
this yields the exponential distribution @xmath67 , as required . for definiteness ,
our numerical calculations were performed with @xmath68 , @xmath69 , @xmath70 and @xmath71 . for the anharmonic potential
all results reported in this paper were obtained for @xmath72 and @xmath73 .
the size of the system ranged from @xmath74 up to @xmath75 .
we also present numerical results for the harmonic case ( @xmath28 ) , for @xmath72 and compare with the results of the anharmonic case .
the existing exact solution @xcite for harmonic case is used to check our numerical procedure .
the exact solution is obtained by solving the equations for the pair correlations , which is possible because they consist of closed equations .
however , this closure property does not happen for the anharmonic case . in fig .
[ vel ] we show the results for the average kinetic energy for each particle as a function of the position @xmath22 on the chain for the harmonic and anharmonic cases .
the temperatures of the reservoirs are considered to be distinct , @xmath76 , and our numerical calculations were performed for @xmath77 and @xmath78 . without stochastic collisions ( @xmath14 ) the results of the harmonic case shows that the kinetic energy is almost constant , a result obtained by rieder et al @xcite , which does not lead to fourier s law .
the inclusion of the stochastic collisions ( @xmath79 ) produces a drastically different result .
the average kinetic energy as a function of @xmath22 displays now a nonzero slope as can be seen in fig .
[ vel ] . for the anharmonic case ,
all curves , including the case @xmath14 , show a nonzero slope . in spite of that
, the @xmath14 case does not lead to fourier s law .
we have calculated the flux @xmath0 at the stationary state by using equation ( [ 15 ] ) and the results are shown in fig .
[ flux ] as a function of @xmath80 for several values of @xmath12 . from this figure
we see clearly that @xmath81 , for sufficiently large values of @xmath4 , in accordance with fourier s law , as long as @xmath79 , for both harmonic and anharmonic cases . for sufficiently large values of @xmath4
the heat conductivity is given by @xmath82 .
this quantity is plotted as a function of @xmath4 for several values of the rate of stochastic collisions @xmath12 , including @xmath14 .
the results are shown in fig .
[ kappa ] .
for both the harmonic and anharmonic cases , the heat conductivity @xmath3 approaches a constant when @xmath83 , as long as @xmath79 , and fourier s law is accomplished .
when @xmath14 , our numerical results gives a superdiffusive behavior with @xmath84 when @xmath83 , according to @xmath85 as shown in fig .
[ kappa ] . for the harmonic case , @xmath18 , which is in accordance with the result by rieder et al @xcite . for the anharmonic case
we get @xmath17 .
this is in agreement with the result @xmath86 found by lepri et al .
@xcite and in excelent agreement with the value @xmath87 @xcite . to analyze the behavior of the heat conductivity @xmath82 as
@xmath88 we have plotted this quantity as a function of @xmath12 for several values of the system size @xmath4 .
the results are shown in fig .
[ kappalambda ] .
we have plotted also the extrapolated values of @xmath3 when @xmath83 for each @xmath12 .
these extrapolated values were extracted from the plot of @xmath89 versus @xmath80 .
the heat conductivity @xmath3 diverges when @xmath88 according to @xmath90 for the harmonic case our results give @xmath19 , a result obtained by dhar et al @xcite . for the anharmonic case we found @xmath15 , a result clearly distinct from the harmonic case .
the algebraic behavior of @xmath3 with @xmath4 and @xmath12 can be obtained by assuming the following finite - size scaling for the heat conductivity , @xmath91 where @xmath92 is a universal function of @xmath93 such that @xmath94 is a finite constant and @xmath95 when @xmath96 is large . to ensure a finite conductivity in the limt @xmath83
, the exponent @xmath97 must be related to the exponents @xmath16 and @xmath98 by @xmath99 . in fig .
[ colaps ] , we show the data collapse for the harmonic case by plotting @xmath100 as a function of @xmath101 , where in this case @xmath18 and @xmath102 .
the data collapse is well described by the expression @xmath103 as can be seen in fig .
[ colaps ] .
this function was obtained by numerically solving the exact equations for the pair correlations from which we found @xmath104 and @xmath105 . for large values of @xmath96 , this result gives @xmath106 so that @xmath107 , independent of @xmath4 , which is the behavior of the conductivity for the harmonic chain when @xmath83 @xcite .
for the anharmonic case , the data colapse , shown in fig .
[ colaps ] , was obtained by using the exponents @xmath16 and @xmath98 obtained previously .
it is worth mentioning that lepri et al .
@xcite in their study of a one - dimensional harmonic crystal with energy conservative noise consisting of elastic collisions between neighboring particles reported a conductivity that behaves as @xmath108 for large values of @xmath4 . assuming for this system a scaling function of the form ( [ 23 ] ) and keeping in mind that @xmath18 , because the conductivity behaves as @xmath109 when @xmath14
, we may conclude that @xmath110 and @xmath102 .
notice , however , that @xmath111 so that @xmath3 is not finite but diverges when @xmath83 .
a relevant feature of the present anharmonic chain with random reversal of velocities is the dependence of the heat conductivity with temperature .
for the harmonic case the heat conductivity is temperature independent @xcite , a result that can be understood by using the following reasoning . if we rescale the temperature of the reservoirs by a factor @xmath112 , that is , @xmath113 and @xmath114 , and the positions and velocities by a factor @xmath115 , that is , @xmath116 and @xmath117 , the equations of motions for the case @xmath28 become invariant .
in addition , according to eq .
( [ 15 ] ) with @xmath28 , the heat flux changes as the temperature , that is , @xmath118 . from this relation
it follows that @xmath0 is a homogeneous function of @xmath33 and @xmath34 , so that @xmath119 .
writing @xmath120 , then for small values of @xmath5 it follows that @xmath121 leading to a heat conductivity @xmath3 independent of temperature , a result that we have also checked numerically . for the anharmonic case we have found that the heat conductivity @xmath3 is an increasing function of temperature . in fig .
[ deptemp ] we show @xmath3 as a function of the temperature @xmath122 of the colder reservoir , the heat conductivity was determined from @xmath82 , for the same value of @xmath123 and for several values of the noise parameter @xmath12 .
we used @xmath124 , a value high enough so that the values of @xmath3 may be considered the asympototic values ( see fig .
[ kappa ] ) , with the exception of the case @xmath14 .
our results indicate a monotonic increase of the heat conductivity with temperature as can be seen in fig .
[ deptemp ] .
we have found that our results are consistent with the result of aoki and kusnezov @xcite and with the upper bounds of bernardin and olla @xcite .
in conclusion , we have considered a chain of particles interacting through anharmonic potentials and subject to heat reservoirs at its ends .
in addition , the chain is subject to a shot noise that changes the sign of the velocities of the particles at random times , distributed according to an exponential distribution .
the shot noise does not change the energy so that the changes in energy are only due to the contact with the reservoirs .
we have shown that , when the chain is connected to reservoirs at different temperatures , the heat flux is inversely proportional to the size of the system , as long as the shot noise parameter @xmath12 is nonzero , and therefore in accordance with fourier s law .
our results suggest , in accordance with @xcite , that the ergodicity may play a crucial role in the derivation of fourier s law .
we have also obtained the behavior of @xmath3 with @xmath4 when @xmath14 and @xmath3 with @xmath12 when @xmath83 .
both behaviors are found to be algebraic , characterized by exponents @xmath125 and @xmath126 .
this allows us to introduce a finite - size scaling in which the noise parameter scales with the inverse of the system size to the power @xmath127 . for the harmonic case
, we have shown that a finite size scaling exists , a result that has been also obtained by numerically solving the equations for the correlations .
we wish to thank e. pereira for his helpful comments .
we acknowledge the brazilian agencies fapesp and cnpq for financial support . | we analyze the transport of heat along a chain of particles interacting through anharmonic potentials consisting of quartic terms in addition to harmonic quadratic terms and subject to heat reservoirs at its ends .
each particle is also subject to an impulsive shot noise with exponentially distributed waiting times whose effect is to change the sign of its velocity , thus conserving the energy of the chain .
we show that the introduction of this energy conserving stochastic noise leads to fourier s law .
the behavior of thels heat conductivity for small intensities of the shot noise and large system sizes are found to obey a finite - size scaling relation .
we also show that the heat conductivity is not constant but is an increasing monotonic function of temperature .
pacs numbers : 05.10.gg , 05.70.ln , 05.60.-k |
geometrical frustration in strongly - correlated systems is one of the long - standing problems in condensed matter physics .
frustration suppresses a formation of a simpleminded long - range order and results in nearly - degenerate ground - state manifolds of a large number of different states .
many well - known examples are found in frustrated antiferromagnetic ( af ) spin systems . there ,
all the antiparallel spin conditions between interacting pairs can not be satisfied at the same time because closed loops contain an odd number of sites .
the degeneracy due to the frustration yields nontrivial phenomena such as complicated ordering structures , spin liquid states , and glassy states .
@xcite besides the spin degree of freedom , charge ordering phenomena are also much affected by the geometrical frustration .
@xcite quantum and thermal fluctuations play important roles in these systems , which are difficult to handle in a controllable manner .
pyrochlore lattice is a typical example of the geometrically - frustrated structures , and it consists of a three - dimensional ( 3d ) network of corner - sharing tetrahedra as shown in fig . [
fig : lattice structure ] ( a ) .
spin systems on the pyrochlore lattice have been intensively studied .
@xcite in particular , for quantum @xmath3 spin systems with only nearest - neighbor interactions , it is predicted that a macroscopic number of singlet states lie inside the singlet - triplet gap .
@xcite several types of symmetry breakings are predicted within the singlet subspace , e.g. , dimer / tetramer ordering , but without magnetic long - range ordering .
@xcite antiferromagnetic ( af ) classical spin systems on the pyrochlore lattice are also believed to show no long - range ordering at any temperature .
@xcite due to the three dimensionality and the large unit cell ( 16 sites in the cubic unit cell ) , pyrochlore systems remain a big challenge to theoreticians and are still far from comprehensive understanding .
we can find many pyrochlore systems in real compounds .
most typically , pyrochlore systems are realized in so - called @xmath4 spinel oxides , where only @xmath4-site cations are magnetic in the general chemical formula @xmath5o@xmath2 .
figure [ fig : lattice structure ] ( b ) shows the @xmath4 spinel structure , which consists of a network of edge - sharing @xmath4o@xmath6 octahedra .
@xmath4 cation is shown by ball in center of each octahedron , and octahedron corners are occupied by oxygen ions , while nonmagnetic @xmath0-site cations are not shown for simplicity . with omitting oxygen ions on the corners of the octahedra , one obtains the pyrochlore lattice of @xmath4 cations in fig .
[ fig : lattice structure ] ( a ) . in this paper
, we will investigate insulating vanadium spinel oxides , @xmath0v@xmath1o@xmath2 with divalent @xmath0-site cations such as zn , mg , or cd . in these compounds , each v@xmath7 cation has two @xmath8 electrons in a high - spin state by the hunds - rule coupling , and they are mott insulators .
@xcite thus , we may consider that pyrochlore spin systems with @xmath9 are realized .
curie - weiss temperature is estimated from the magnetic susceptibility as @xmath10k , and any long - range ordering does not occur down to significantly lower temperatures than @xmath11 due to the geometrical frustration .
@xcite for instance , in znv@xmath1o@xmath2 , a structural phase transition occurs at @xmath12k from the high - temperature cubic phase to the low - temperature tetragonal phase with a flattening of vo@xmath6 octahedra in the @xmath13 direction .
@xcite successively , an af transition occurs at @xmath14k .
@xcite neutron scattering experiments revealed that the af ordering structure below @xmath15 is a collinear one which consists of the staggered af chains in the @xmath16 planes stacking in the @xmath13 direction with a four - times period as up - up - down - down-@xmath17 as shown in fig .
[ fig : spin order ] .
@xcite these two successive transitions are commonly seen in compounds mgv@xmath1o@xmath2 and cdv@xmath1o@xmath2 .
@xcite this indicates that the degenerate ground - state manifolds in the pyrochlore systems are lifted at low temperatures in some manner .
a few years ago , yamashita and ueda proposed a scenario to explain the mechanism of the transitions in @xmath0v@xmath1o@xmath2 .
@xcite their approach is based on a valence - bond - solid picture for @xmath9 spins and takes account of the coupling to jahn - teller ( jt ) lattice distortions .
they claimed that the first transition at @xmath18 is due to the jt effect which lifts the degeneracy of the spin - singlet local ground states at each tetrahedron unit .
this scenario based on the spin - jt coupling is appealing , however , some difficulty still remains .
the problem is that it is difficult to explain the magnetic transition at a lower temperature @xmath15 . in this approach
, a finite energy gap is assumed between the spin - singlet ground - state subspace and the spin - triplet excitations , and a low - energy effective theory is derived to describe a phase transition within the spin - singlet subspace .
high - energy excitations with total spin @xmath19 are already traced out at the starting point , and therefore their effective model has no chance to describe the af ordering within their theory .
a similar jt scenario was also examined for classical spin systems .
@xcite in this case , although the problem to have af order does not exist , there is another difficulty to explain the following generic difference from chromium family of @xmath0cr@xmath1o@xmath2 ( @xmath0=zn , mg , or cd ) .
these chromium oxides are also @xmath4 spinels and magnetic cr cations constitute a pyrochlore lattice .
however , in contrast to the two transitions in vanadium compounds , the chromium compounds exhibit only one transition , i.e. , the af order appears simultaneously with the structural transition .
@xcite this clear difference is generic , being independent of divalent @xmath0 cations , and ascribed to the difference of magnetic cations v@xmath7 and cr@xmath7 , which can not be explained by the classical spin approach based on the spin - jt effect unless there exists essential difference in the model parameters between the two families . therefore
, these spin - jt type theories appear to be insufficient to explain the mechanism of two transitions in vanadium spinels @xmath0v@xmath1o@xmath2 .
these insulating compounds are undoped states of liv@xmath1o@xmath2 which exhibits a unique heavy fermion behavior .
@xcite the origin of the mass enhancement is still controversial between the scenario based on the kondo effect @xcite and the scenario of strong correlations with the geometrical frustration .
@xcite since the doping of li shows systematic changes of magnetic @xcite and transport properties , @xcite as well as the phase diagram , @xcite understanding of undoped materials may give a starting point to discuss the doped state in liv@xmath1o@xmath2 .
therefore , it is also highly desired to clarify the mechanism of orderings in the undoped compounds @xmath0v@xmath1o@xmath2 .
the generic difference between vanadium and chromium spinels mentioned above suggests an importance of @xmath20 orbital degrees of freedom . in the case of chromium spinels ,
each cr@xmath7 cation has three electrons in threefold @xmath20 levels and large hunds - rule coupling leads to a high - spin state , and therefore , there is no orbital degree of freedom . on the contrary , in the case of vanadium spinels
, since each v@xmath7 cation has two electrons , the orbital degree of freedom is active . with taking account of this @xmath20 orbital degeneracy
, the authors have derived an effective spin - orbital - lattice coupled model in the strong correlation limit and investigated it by mean - field type arguments .
@xcite a reasonable scenario was obtained , but discussions were limited to a qualitative level . in order to investigate temperature dependences of physical properties semiquantitatively accurate enough to be compared with experimental data , we need more elaborate analysis . in the present study , we will investigate thermodynamic properties of the effective spin - orbital - lattice model derived by the authors in ref . by extensive monte carlo ( mc ) calculations .
we will show that this model indeed exhibits two successive transitions in a reasonable parameter range , and clarify the microscopic mechanism of these transitions in detail .
first , an orbital order appears with the tetragonal jt distortion which flattens vo@xmath6 octahedra .
this orbital order reduces magnetic frustration partially , and enhances af spin correlations in one - dimensional ( 1d ) chains in the @xmath16 planes . at a lower temperature ,
the third - neighbor exchange interaction and thermal fluctuations align these 1d af chains coherently , and stabilize a 3d collinear af order . with comparing numerical results of physical quantities with experimental data , we will show that our theory captures essential physics at low temperatures in the vanadium spinels @xmath0v@xmath1o@xmath2 .
this paper is organized as follows . in sec .
[ sec : model and method ] , we introduce the effective spin - orbital - lattice coupled model , and briefly summarize the mean - field arguments discussed in ref . .
realistic parameter values and mc method are also described . in sec .
[ sec : results ] , we show numerical results in comparison with experimental data .
we then make several remarks , in particular , on comparisons with experimental results and other theoretical proposals in sec .
[ sec : discussions ] .
section [ sec : summary ] is devoted to summary .
in the present study , we will investigate thermodynamic properties of the spin - orbital - lattice coupled model which is proposed by the authors in ref . .
the hamiltonian consists of two terms as @xmath21 the first term describes exchange interactions in spin and orbital degrees of freedom and the second term is for orbital - lattice couplings of the jahn - teller type . the spin - orbital hamiltonian @xmath22 is derived from a multiorbital hubbard model with threefold @xmath20 orbital degeneracy by the perturbation in the strong correlation limit .
@xcite the starting @xmath20 hubbard model is given in the form @xmath23 \nonumber \\ & & \quad + \frac12 \sum_i \sum_{\alpha\beta,\alpha'\beta ' } \sum_{\tau\tau ' } u_{\alpha\beta,\alpha'\beta ' } c_{i\alpha\tau}^{\dagger } c_{i\beta\tau'}^{\dagger } c_{i\beta'\tau ' } c_{i\alpha'\tau } , \label{eq : h_multiorbital}\end{aligned}\ ] ] where @xmath24 and @xmath25 are site and spin indices , respectively , and @xmath26 ( @xmath27 ) , @xmath28 ( @xmath29 ) , @xmath30 ( @xmath31 ) are orbital indices .
the first term of @xmath32 is the electron hopping , and the second term describes coulomb interactions , for which we use the standard parametrizations , @xcite @xmath33 we do not include here the relativistic spin - orbit coupling and the trigonal distortion in the model ( [ eq : h_multiorbital ] ) .
effects of these neglected elements will be discussed in sec .
[ sec : symmetry ] and e. considering the vanadium spinel oxides are insulators , it is reasonable to start from the strong correlation limit and treat the hopping term as perturbation .
the unperturbed states are atomic eigenstates with two electrons on each v cation in a high - spin state . as for the perturbation part , on the basis of the tight - binding fit for the band structure @xcite ( see sec .
[ sec : parameters ] ) , we take account of hopping integrals of @xmath34 bonds for nearest - neighbor pairs , @xmath35 , and for third - neighbor pairs , @xmath36 , as shown in fig . [
fig : transfer ] . note that the third - neighbor pair corresponds to the next nearest - neighbor pair along each chain .
the hopping integral between second - neighbor pairs ( the gray arrow in fig .
[ fig : transfer ] ) is expected to be small because of the geometry of the pyrochlore lattice , @xcite and moreover the exchange interaction derived from it is frustrated .
the second order perturbation in @xmath35 and @xmath36 gives the hamiltonian in the form @xmath37 , \label{eq : h_nn } \\ & & h_{\rm so}^{\rm 3rd } = -j_3 \sum_{\langle\!\langle i , j \rangle\!\rangle } \ [ \ h_{\rm o - af}^{(ij ) } + h_{\rm o - f}^{(ij ) } \ ] , \label{eq : h_3rd } \\ & & h_{\rm o - af}^{(ij ) } = ( a + b { \mbox{\boldmath $ s$}}_i \cdot { \mbox{\boldmath $ s$}}_j ) \nonumber \\ & & \quad \times \big [ n_{i \alpha(ij ) } ( 1 - n_{j \alpha(ij ) } ) + ( 1 - n_{i \alpha(ij ) } ) n_{j \alpha(ij ) } \big ] , \label{eq : h_o - af } \\ & & h_{\rm o - f}^{(ij ) } = c ( 1 - { \mbox{\boldmath $ s$}}_i \cdot { \mbox{\boldmath $ s$}}_j ) n_{i \alpha(ij ) } n_{j \alpha(ij ) } , \label{eq : h_o - f}\end{aligned}\ ] ] where @xmath38 is the @xmath9 spin operator and @xmath39 is the density operator for site @xmath40 and orbital @xmath41
. the summations with @xmath42 and @xmath43 are taken over the nearest - neighbor sites and third - neighbor sites , respectively . here
, @xmath44 is the orbital which has a finite hopping integral between the sites @xmath40 and @xmath45 , for instance , @xmath46 ( @xmath31 ) for @xmath40 and @xmath45 sites in the same @xmath47 plane .
the other parameters in eqs .
( [ eq : h_nn])-([eq : h_o - f ] ) are determined by coupling constants in eq .
( [ eq : h_multiorbital ] ) as @xcite @xmath48 and each site is subject to the local constraint , @xmath49 .
realistic values of these parameters are given in sec .
[ sec : parameters ] . an important feature of @xmath22 is the highly anisotropic form of the orbital intersite interaction .
it is a three - state clock type interaction corresponding to three different orbital states , in which there is no quantum fluctuation since the density operator @xmath50 is a constant of motion .
this anisotropy comes from the orbital diagonal nature of the @xmath34-bond hopping integrals which do not mix different orbitals .
moreover , the orbital interaction depends on both the bond direction and the orbital states in two sites . on the other hand ,
the spin exchange interaction is heisenberg type and isotropic , independent of the bond direction . the orbital - lattice term @xmath51 in eq .
( [ eq : h ] ) reads @xmath52 where @xmath53 is the electron - phonon coupling constant and @xmath54 denotes the amplitude of local lattice distortion at site @xmath40 . here , we take account of only the tetragonal mode in the @xmath55 direction . note that this simplification breaks the cubic symmetry of the system .
we choose the sign of @xmath54 such that it is positive for a flattening of vo@xmath6 octahedra which leads to the level splitting in fig .
[ fig : orbital order ] ( a ) .
the second term in eq .
( [ eq : h_jt ] ) denotes the local elastic energy of distortions .
finally , the third term denotes the interaction of jt distortions between nearest - neighbor sites which mimics the cooperative aspect of the jt distortion .
it is reasonable to assume a positive value of @xmath56 because a tetragonal distortion of a vo@xmath6 octahedron modifies its neighboring octahedra in a similar distortion due to the edge - sharing 3d network of octahedra in fig .
[ fig : lattice structure ] ( b ) . for simplicity , we here neglect quantum nature of phonons .
although jt distortions modify @xmath22 through changes of hopping integrals , we neglect these corrections in the present study .
we normalized the variable @xmath54 to absorb the elastic constant in the second term in eq .
( [ eq : h_jt ] ) , hence the dimension of @xmath54 is ( energy)@xmath57 . as a result , the dimensions of @xmath53 and @xmath56 are ( energy)@xmath57 and ( energy ) , respectively .
for the following discussions , we here briefly summarize the results of the mean - field type analysis on the model ( [ eq : h ] ) obtained in ref . .
the analysis predicts that first , the degeneracy due to the geometrical frustration will be partially lifted in the orbital channel .
there , the anisotropy of the orbital interaction in @xmath58 plays a crucial role ; the interaction is a three - state clock type and depends on the orbital states as well as the bond direction .
the remaining degeneracy of orbital states is lifted by the tetragonal jt coupling in @xmath51 : a flattening of vo@xmath6 octahedra splits the threefold orbital levels as shown in fig .
[ fig : orbital order ] ( a ) , and selects the orbital ordering structure as shown in fig .
[ fig : orbital order ] ( b ) . there , one of the two electrons occupies the @xmath31 orbital at every site , and the other occupies either @xmath27 or @xmath29 orbital in an alternative manner along the @xmath55 direction .
when we consider only the nearest - neighbor interactions @xmath58 , this orbital occupation induces af spin interactions on the bonds within the @xmath16 planes [ the blue solid lines in fig .
[ fig : orbital order ] ( b ) ] and ferromagnetic spin interactions on the bonds among the @xmath16 planes [ the red dashed lines in fig .
[ fig : orbital order ] ( b ) ] .
the coupling constant for the former af interaction is @xmath59 and that for the latter ferromagnetic interaction is @xmath60 . since @xmath61 is a small parameter of the order of 0.1 as estimated in sec .
[ sec : parameters ] , the former af interaction is much larger than the latter ferromagnetic one .
moreover , the ferromagnetic interactions are frustrated for the af spin configuration within the @xmath16 planes because of the geometry of the pyrochlore lattice . hence , under the orbital ordering shown in fig .
[ fig : orbital order ] ( b ) , the af spin correlations develop within the 1d chains in the @xmath16 planes , and the 1d af chains are independent with each other ; relative angles among the af moments are not yet determined at this stage .
the relative angles are partially fixed by including the third - neighbor interactions @xmath62 : the third - neighbor interactions align the af moments in two next - neighboring @xmath16 planes , and lead to a 3d collinear af order with the wave vector @xmath63 .
( see fig .
[ fig : two sublattices ] . ) in the ordered state , however , there are two independent af sublattices ; one consists of @xmath64 $ ] chains and the other consists of @xmath65 $ ] chains .
the relative angle between the af moments on the two sublattices is still free in this mean - field argument . from the spin wave calculation of the zero - point energy
, we discussed that quantum fluctuations fix the relative angle and stabilize a collinear af order in fig .
[ fig : spin order ] that is consistent with the neutron scattering result .
the mean - field argument gives a reasonable scenario for two transitions in @xmath0v@xmath1o@xmath2 , but the argument is limited to a qualitative level . in order to confirm the scenario and understand the experimental results more quantitatively , we need more sophisticated analysis , especially for the thermodynamic properties of the system . in the present study , we will perform the monte carlo simulation for this purpose . here
, we estimate realistic values of parameters in the model ( [ eq : h ] ) which are given by the parameters in the starting @xmath20 hubbard model ( [ eq : h_multiorbital ] ) .
as for the hopping parameters @xmath66 in @xmath32 , a tight - binding fit to the results of the first - principle band calculation suggests the dominant hopping integrals are those of @xmath34 bonds , and gives @xmath67ev and @xmath68ev for nearest - neighbor and third - neighbor pair of sites , respectively .
@xcite for coulomb interactions , there are estimates based on the cluster analysis for optical experiments for vanadium perovskites @xmath0vo@xmath69 , which also have a vo@xmath6 octahedral unit .
@xcite the estimates are @xmath70ev and @xmath71ev , thus , @xmath72 in eq .
( [ eq : eta ] ) is a small parameter of the order of 0.1 .
we will set @xmath73 in the following numerical calculations .
the estimates of @xmath35 , @xmath36 , and @xmath74 give @xmath75k and @xmath76k , i.e. , @xmath77 . in the monte carlo calculations , we study mainly the case of @xmath78 , but we vary the value of @xmath79 from 0 to 0.05 to examine the systematic change by @xmath80 . hereafter , we will set @xmath81 as an energy unit and the lattice constant of cubic unit cell as a length unit ( @xmath82 ) , and use the convention of the boltzmann constant @xmath83 .
it is hard to estimate the electron - phonon interaction parameters @xmath53 and @xmath56 , and therefore , we treat them as variable parameters in the present study . in the present mc calculations to confirm the above mean - field scenario , we are interested in the parameter region where the orbital order in fig . [
fig : orbital order ] ( b ) is stabilized by the tetragonal jt distortion with a flattening of vo@xmath6 octahedra .
the stability conditions for this orbital and lattice order will be obtained in appendix a by a mean - field type argument . in the following mc calculations
, we will show mc results for typical values of @xmath53 and @xmath56 which satisfy the conditions as @xmath84 and @xmath85 . in the present study
, we will use mc calculations to investigate thermodynamic properties of the effective spin - orbital - lattice model ( [ eq : h ] ) .
quantum mc simulations for frustrated systems are known to be difficult because of the negative sign problem . in the present mc study ,
we neglect quantum fluctuations and approximate the model in the classical level .
this approximation retains effects of thermal fluctuations which may play dominant roles in finite - temperature transitions .
the quantum nature originates only from spin @xmath9 operators in the model ( [ eq : h ] ) , since the orbital interaction is classical and is a diagonal one of three - state clock type and since jt distortions are also treated as classical variables .
thus , we approximate the spin operators by classical vectors with the modulus @xmath86 [ the length of the vector is normalized to give the same largest @xmath55 component @xmath87 ( classical part ) ]
. effects of quantum fluctuations will be discussed by using the spin wave approximation in sec .
[ sec : quantum fluctuation ] .
thereby , the model ( [ eq : h ] ) consists of the classical heisenberg part for spins , the three - state clock part for orbitals , and the classical phonon part .
we use a standard metropolis mc algorithm . in the actual mc calculations ,
the mc sampling is performed to measure spin vectors @xmath38 , three - state clock spins for the orbital states ( defined in sec .
[ sec : orbital ordering ] ) , and amplitudes of the jt distortion @xmath54 at all the lattice sites .
we typically perform @xmath88 mc samplings for measurements after @xmath88 steps for thermalization .
the measurements are performed in every @xmath89-times mc update , and we typically take @xmath90 .
results are divided into five bins to estimate statistical errors by variance of average values in the bins . here , one mc update consists of several - times sweeps ( typically twice ) for spin directions , orbital states , and jt distortions .
the one sweep is @xmath91-times trials by choosing a site randomly , where @xmath91 is the number of lattice sites . in the sampling on spin directions ,
we apply the so - called pivot rotation when a trial is rejected , which is a precession without energy cost : the pivot rotation of a spin is achieved by a random rotation with keeping the relative angle to the mean - field vector determined by its nearest - neighbor and third - neighbor spin and orbital states .
this accelerates the mc sampling in the configurational space .
as shown in sec .
[ sec : results ] , the orbital transition accompanied by the jt distortion is first order . to avoid a hysteresis and determine the transition temperature precisely , we start mc calculations at each temperature from a mixed initial condition for orbital and lattice states in which a half of the system takes a low - temperature ordered configuration and the rest takes a high - temperature disordered configuration
this technique is known to be free from trapping at a metastable state for large enough system sizes .
@xcite at very low temperatures , we use an perfectly ordered initial state to accelerate the convergence .
the system sizes in the present work are up to @xmath92 where @xmath93 is the linear dimension of the system measured in the cubic units , i.e. , the total number of sites @xmath91 is given by @xmath94 .
in this section , we present mc results for the model ( [ eq : h ] ) in comparison with experimental data . in secs . [
sec : transitions]-c , we show the results for the typical case of @xmath78 .
systematic changes with @xmath79 and a generic phase diagram will be discussed in sec .
[ sec : phase diagram ] . in the following
, we present mc results for @xmath78 to show that the model ( [ eq : h ] ) exhibits two phase transitions with temperature . in sec .
[ sec : e and c ] , we present the mc results for the internal energy and the specific heat , which show two different anomalies . we also discuss the changes of the entropy related to the two transitions . in secs .
[ sec : orbital ordering ] and 3 , we discuss the nature of the two transitions by calculating the order parameters . in sec .
[ sec : collinearity ] , we examine the magnetic ordering structure in the low temperature phase , and point out the importance of thermal fluctuations . sec .
[ sec : chi ] contains mc results for the uniform magnetic susceptibility for comparison with experimental data .
figure [ fig : e and c ] shows temperature dependences of the internal energy and the specific heat per site .
the internal energy per site is calculated by the thermal average of the hamiltonian ( [ eq : h ] ) as @xmath95 the specific heat is calculated by fluctuations of the internal energy as @xmath96 as shown in fig .
[ fig : e and c ] ( a ) , the internal energy @xmath97 jumps at @xmath98 .
it indicates that a first - order transition occurs at this temperature .
the jump is also found in the specific heat at the same temperature in fig .
[ fig : e and c ] ( b ) .
the specific heat shows another anomaly at a lower temperature @xmath99 .
there , we find a systematic enhancement of the peak as the system size increases , which is a sign of second - order phase transition .
thus , mc data in fig .
[ fig : e and c ] indicate that the system shows two different transitions ; the first - order transition at @xmath100 and the second - order transition at @xmath101 . in the following sections , the two transitions are to be assigned to the orbital ordering with tetragonal lattice distortion and the af spin ordering , respectively . ) ] and ( b ) the specific heat per site [ eq .
( [ eq : c ] ) ] at @xmath78 .
error bars are smaller than symbol sizes in ( a ) .
typical error bars are shown in ( b ) . , width=264 ] it is noted that the specific heat approaches @xmath102 ( in units of @xmath103 ) as @xmath104 as shown in fig .
[ fig : e and c ] ( b ) .
a finite value of @xmath105 at @xmath106 is characteristic to classical models , and one degree of freedom remaining at the ground state contributes @xmath107 to @xmath105 .
hence , the data in fig .
[ fig : e and c ] ( b ) suggests that there remain three degrees of freedom at @xmath106 in the present model . from the discussions of the ordered phase at low temperatures in the following sections
, they can be ascribed to two transverse modes of the af spin order and one jt mode . from the jump of the internal energy in fig .
[ fig : e and c ] ( a ) , we estimate the entropy jump @xmath108 associated with the first - order transition from the disordered phase above @xmath109 to the ordered phase below @xmath109 .
the two phases have the same free energy @xmath110 at @xmath109 , which gives us the entropy difference @xmath108 as @xmath111 this corresponds to @xmath112-@xmath113% of @xmath114 per site .
the amount of the entropy which is related to the fluctuation around the second - order transition at @xmath115 is estimated from the area of the anomalous peak at @xmath115 in the plot of @xmath116 as a function of @xmath117 .
although it is difficult to estimate it because of the large system - size dependence of @xmath105 as well as the large error bars in fig .
[ fig : e and c ] ( b ) , a rough estimate is obtained by an interpolation with polynomial functions for the normal contribution and a numerical integration of the anomalous part .
the estimate is roughly @xmath118-@xmath119 , which corresponds to @xmath120-@xmath121% of @xmath114 per site . in experiments ,
the amounts of entropy related to the two transitions are also estimated from the specific heat . in znv@xmath1o@xmath2 ,
the entropy change in the higher - temperature transition at @xmath18 is @xmath122-@xmath123j / mol k , i.e. , @xmath124% of @xmath114 per v cation .
@xcite the entropy related to the lower - temperature transition at @xmath15 is small and not estimated quantitatively , but roughly less than @xmath125j / mol k , i.e. , @xmath126% of @xmath114 per v cation .
@xcite in mgv@xmath1o@xmath2 , the former is estimated as @xmath122j / mol k , i.e. , @xmath127% of @xmath114 , and the latter is @xmath128j / mol k , i.e. , @xmath129% of @xmath114 per v cation . @xcite our estimates from the mc results in fig .
[ fig : e and c ] show semiquantitative agreement with these experimental values , and particularly , explain that the entropy related to the lower - temperature transition is considerably smaller than that for the higher - temperature transition .
the small entropy related to the transition at @xmath115 is likely due to the magnetic frustration and a 1d af correlation well developed above @xmath115 which will be discussed in sec .
[ sec:1d ] . to characterize two transitions in fig .
[ fig : e and c ] , we calculate corresponding order parameters .
first , we consider the transition at the higher temperature @xmath100 .
figure [ fig : orbital ] ( a ) shows the sublattice orbital moment , which is defined in the form @xmath130 where the summation is taken over the sites within one of the four sublattices in fig .
[ fig : orbital order ] ( b ) . here , @xmath131 is the three - state clock vector at the site @xmath40 which describes three different orbital states as shown in the inset of fig .
[ fig : orbital ] ( a ) ; @xmath132 for ( @xmath133 ) , @xmath134 for ( @xmath135 ) , and @xmath136 for ( @xmath137 ) orbital occupations , respectively .
it is found that the values of @xmath138 for four different sublattices have the same value within the error bars so that we omit the sublattice index in eq .
( [ eq : m_o ] ) . as shown in fig .
[ fig : orbital ] ( a ) , @xmath138 shows a clear jump at the same temperature as for the internal energy and the specific heat .
this suggests that a four - sublattice orbital ordering occurs at @xmath109 . at low temperatures ,
@xmath138 approaches its maximum value @xmath125 , which indicates the four - sublattice orbital order becomes almost perfect there .
figures [ fig : orbital ] ( b)-(e ) show the orbital distribution for four sublattices @xmath125-@xmath123 shown in fig .
[ fig : orbital order ] ( b ) , respectively , which is defined as @xmath139 where @xmath140 ( @xmath27 ) , @xmath28 ( @xmath29 ) , and @xmath30 ( @xmath31 ) .
the results indicate that at @xmath109 the orbital distributions suddenly change from equally distributed @xmath141 in the para phase above @xmath109 to almost polarized @xmath142 or @xmath125 for @xmath143 . in the orbital ordered phase below @xmath109 , @xmath27 ( @xmath144 ) and @xmath31 ( @xmath145 ) orbitals
are occupied in the sublattices 1 and 4 , and @xmath29 ( @xmath146 ) and @xmath31 ( @xmath145 ) orbitals are occupied in the sublattices 2 and 3 ( ref . ) .
this orbital ordering structure is shown in fig . [
fig : orbital structure ] .
this pattern is consistent with the mean - field prediction in fig .
[ fig : orbital order ] ( b ) . ) at @xmath78 .
the inset shows the three - state clock vector for the orbital state .
( b)-(d ) electron density in each orbital for four sublattices @xmath125-@xmath123 shown in fig .
[ fig : orbital order ] ( b ) for @xmath147 [ eq . ( [ eq : nbar ] ) ] .
error bars are smaller than the symbol sizes .
, width=302 ] accompanying the transition at @xmath109 , a tetragonal jt distortion occurs discontinuously . in fig .
[ fig : jt ] , we plot the average of the jt distortions which is calculated by @xmath148 the positive value of @xmath149 below @xmath109 corresponds to a ferro - type tetragonal jt distortion with a flattening of vo@xmath6 octahedra as mentioned in sec .
[ sec : spin - orbital model ] .
therefore , the level splitting shown in fig .
[ fig : orbital order ] ( a ) is realized .
the value of @xmath149 approaches 2 at low temperatures which is the mean - field value obtained in appendix a. we note that @xmath149 is small but finite even for @xmath150 .
this is because @xmath51 breaks the cubic symmetry of the system as mentioned before . ) at @xmath78 .
error bars are smaller than the symbol sizes . ,
width=264 ] therefore , the discontinuous phase transition at @xmath109 is ascribed to the orbital ordering with the pattern of fig .
[ fig : orbital structure ] accompanied by the tetragonal jt distortion with the flattening of vo@xmath6 octahedra . from figs . [ fig : orbital ] and [ fig : jt ] , we estimate @xmath151 for @xmath152 .
next , we consider the other transition at @xmath101 .
figure [ fig : spin ] ( a ) shows the temperature dependence of the staggered magnetization defined in the form @xmath153 where the structure factor is given by @xmath154 .
\label{eq : f}\ ] ] this definition looks complicated but nothing but the order parameter of the spin ordering pattern shown in fig .
[ fig : spin order ] .
here , the first summation is taken over different chains lying in the @xmath47 planes ( @xmath155 is the total number of the @xmath47 chains in the system , i.e. , @xmath156 where @xmath93 is the linear dimension of the system measured in the cubic units ) , and the second summation @xmath157 is taken over the sites in the @xmath158-th @xmath47 chain .
@xmath159 is the @xmath160 coordinates of the site @xmath40 , and @xmath161 is the @xmath55 coordinate of the @xmath158-th @xmath47 chain measured in the cubic units .
we set the normalization in eq .
( [ eq : f ] ) such that @xmath162 becomes @xmath125 for the fully saturated af order in fig .
[ fig : spin order ] .
note that the structure factor @xmath163 can not be defined only by a real phase factor because of the complicated af ordering pattern which is expected to have the four - period structure in the @xmath164 and @xmath165 directions as shown in fig .
[ fig : spin order ] .
see also the discussions in sec .
[ sec : collinearity ] .
the structure factor is also expressed in a simpler form as @xmath166 where the form factor @xmath167 is given by @xmath168 + i \cos[2\pi ( x_i - y_i)].\ ] ] note that @xmath167 is specified only by the @xmath169 and @xmath160 coordinates of the site @xmath40 because the @xmath55 coordinate is uniquely determined within the cubic unit cell due to the special structure of the pyrochlore lattice .
[ see the projection in fig . [
fig : spin order ] ( b ) . ] as shown in fig .
[ fig : spin ] ( a ) , the staggered magnetization @xmath162 develops continuously below @xmath101 , and approaches the fully saturated value @xmath170 as @xmath104 .
figure [ fig : spin ] ( b ) shows the staggered magnetic susceptibility obtained by @xmath171 here , we calculate the susceptibility by using @xmath172 instead of @xmath173 in eq .
( [ eq : chi_s ] ) for convenience of numerical calculations since both quantities agree with each other in the thermodynamic limit .
the susceptibility @xmath174 shows a diverging behavior at @xmath115 and the peak value increases with the system size . in fig .
[ fig : spin ] ( c ) , we show the binder parameter for the staggered magnetization which is defined as @xcite @xmath175 it is known that the binder parameter becomes larger ( smaller ) for larger system sizes in the ordered ( disordered ) phase , and hence , the crossing point of the binder parameter for different system sizes gives a good estimate of the transition temperature . the mc data in fig .
[ fig : spin ] ( c ) shows the crossing at the same temperature of the divergence of @xmath174 in fig .
[ fig : spin ] ( b ) , which indicates the phase transition by the order parameter @xmath162 at @xmath115 . ) .
error bars are smaller than the symbol sizes .
( b ) the staggered magnetic susceptibility in eq .
( [ eq : chi_s ] ) .
( c ) the binder parameter defined in eq .
( [ eq : g_s ] ) .
the lines are guides for the eyes .
all the results are calculated at @xmath78 .
, width=264 ] all these results in fig .
[ fig : spin ] indicate that the phase transition at @xmath115 is caused by the continuous growth of the staggered magnetic order . from figs .
[ fig : spin ] ( b ) and ( c ) , we estimate @xmath176 for @xmath152 .
the magnetic ordering pattern will be discussed in the next section . the mean - field argument in ref . and in sec .
[ sec : spin - orbital model ] predicts that the staggered magnetizations develop independently on two sublattices and the relative angle between two af moments is free .
the two sublattices are shown in fig .
[ fig : two sublattices ] ; one consists of the @xmath64 $ ] chains and the other consists of the @xmath65 $ ] chains .
the staggered magnetization @xmath162 defined in eq .
( [ eq : m_s ] ) is the order parameter of the 3d af spin ordering in fig .
[ fig : spin order ] , but can not measure the collinearity between two staggered magnetizations .
for example , @xmath162 takes the value of @xmath125 independent of the relative angle between staggered moments in two sublattices , @xmath177 and @xmath178 , shown in fig .
[ fig : two sublattices ] , when both sublattice moments saturate .
the spin configuration in fig .
[ fig : two sublattices ] is a noncollinear one , while that in fig .
[ fig : spin order ] is a collinear one , and both give @xmath170 . here
, we measure the collinearity between the staggered moments in the two sublattices by @xmath179 here we may consider @xmath180 the angle between @xmath177 and @xmath178 . in the present calculations , the two sublattice moments are obtained by the real and imaginary parts of the structure factor @xmath163 in eq .
( [ eq : f ] ) as @xmath181 respectively . if the af order is collinear , i.e. , @xmath182 , the value of @xmath183 become @xmath125
. figure [ fig : c12 ] shows the mc results . below @xmath101 , @xmath183 is rapidly enhanced , and becomes larger and approaches @xmath125 as the system size increases .
this indicates that the magnetic state is collinear below @xmath115 .
thus , the af order below @xmath115 is identified by the collinear af spin order whose pattern is given by fig .
[ fig : spin order ] . ) at @xmath78 .
the lines are guides for the eyes .
, width=264 ] the result indicates that thermal fluctuations in the classical version of the model ( [ eq : h ] ) lift the degeneracy of the relative angle between two sublattice magnetizations and stabilize the collinear spin structure .
this is a kind of the so - called order - by - disorder phenomena .
@xcite in sec .
[ sec : spin - orbital model ] and ref . , we discussed that the collinear order may appear also by quantum fluctuations .
thus , we can conclude that both thermal and quantum fluctuations favor a collinear state of the two sublattice magnetizations .
it is known that in many frustrated systems , thermal and quantum fluctuations favor a collinear state .
this is also the case for the model ( [ eq : h ] ) .
we show in fig .
[ fig : chi ] the temperature dependence of the uniform magnetic susceptibility calculated by @xmath184 where @xmath185 is the total magnetic moment per site , @xmath186 the result corresponds to the zero - field - cool ( zfc ) result in experiments rather than the field - cool ( fc ) result since the susceptibility is measured by starting the mc simulation from an initial state with zero magnetic field at each temperature as described in sec .
[ sec : monte carlo ] .
the mc results show a sudden drop at @xmath100 and a little continuous change at @xmath101 .
these features qualitatively agree with the experimental results .
@xcite ) at @xmath78 .
error bars are smaller than the symbol sizes . , width=264 ] in experiments , a large difference between zfc and fc results has been found .
the difference develops from well above @xmath18 , and remains substantial even below @xmath15 .
we will comment on this behavior in sec .
[ sec : zfc and fc ] . as shown in fig .
[ fig : spin ] ( a ) , @xmath162 in finite - size systems are enhanced below @xmath109 even above @xmath115 . in this section ,
we show that there the short - range af correlation is much enhanced along 1d chains in the @xmath47 planes compared to the @xmath164 and @xmath165 chains
. chains [ eq . ( [ eq : m_s^xy ] ) ] and ( b ) the @xmath164 chains [ eq . ( [ eq : m_s^yz ] ) ] at @xmath78 .
error bars are smaller than the symbol sizes . , width=264 ] the mean - field arguments in sec .
[ sec : spin - orbital model ] and in ref . suggest that the orbital order enhances 1d af correlations within the @xmath47 chains in the intermediate phase @xmath187 . to examine the spatially anisotropic spin correlation
, we calculate the staggered magnetic moments along the chains in different directions .
the staggered moment along the @xmath47 chains may be defined by @xmath188^{1/2 } \big\rangle , \label{eq : m_s^xy}\ ] ] where the summations are taken in the same manner as in eq .
( [ eq : f ] ) , and the phase factor describes the @xmath189-@xmath190-@xmath189-@xmath190- @xmath17 structure in the @xmath47 chains as shown in fig .
[ fig : spin order ] .
similarly , the staggered " moment along the @xmath164 chains is calculated by @xmath191^{1/2 } \big\rangle , \label{eq : m_s^yz}\ ] ] where the first summation is taken over all the @xmath164 chains in the system and the second summation is taken over the sites within the @xmath158-th @xmath164 chain with a phase factor of period four in the @xmath55 direction with considering the @xmath189-@xmath189-@xmath190-@xmath190- @xmath17 structure as shown in fig . [ fig : spin order ] . here , @xmath155 is the number of the @xmath47 or @xmath164 chains , i.e. , @xmath192 where @xmath93 is the linear dimension of the system measured in the unit cell .
the normalization factors in eqs .
( [ eq : m_s^xy ] ) and ( [ eq : m_s^yz ] ) are given such that both @xmath193 and @xmath194 become @xmath125 in the fully saturated af state with @xmath195 in eq .
( [ eq : m1&m2 ] ) .
figure [ fig : m_s chains vs t ] shows the mc results .
the moment in the @xmath165 chains , @xmath196 , has the same value as @xmath194 within the statistical error bars .
below @xmath100 , the moment in the @xmath47 chains , @xmath193 , is much more enhanced compared to that in the @xmath164 and @xmath165 chains , @xmath194 and @xmath196 .
this indicates that in the intermediate phase @xmath187 , the af spin correlations develop mainly within the @xmath47 chains .
chains and ( b ) the @xmath164 chains at @xmath78 .
error bars are smaller than the symbol sizes .
the gray lines are the fits to a linear function of @xmath197 .
see the text for details .
, width=302 ] the system - size dependence of these moments within the chains provides further information .
figure [ fig : m_s chain vs l^-0.5 ] plots the system - size extrapolations of the mc data in fig .
[ fig : m_s chains vs t ] . in the ordered phase below @xmath101
, the mc data are extrapolated to finite values , which indicates a long - range order .
the extrapolated values of @xmath193 and @xmath194 to @xmath198 agree with each other within the error bars as expected in this 3d ordered phase . on the other hand , in the disordered phase above @xmath100 , the mc data for both @xmath193 and @xmath194 well scale to @xmath197 and are extrapolated to zero .
( only the data at @xmath199 are shown in the figures as a typical example but other data at @xmath150 show a similar behavior . )
this scaling implies the exponential decay of the spin - spin correlation along the chain as explained below .
compared to the rather isotropic behavior for @xmath200 and @xmath150 , we find a contrasting behavior between @xmath193 and @xmath194 in the intermediate phase @xmath187 ; almost all the data of @xmath194 well scale to @xmath197 even for the smallest system size in the present calculations with @xmath201 , while the data of @xmath193 show a clear deviation from the scaling in the small-@xmath93 region . in order to analyze this contrasting behavior in the intermediate phase
, we introduce the correlation length @xmath202 along the chain , and assume a simple scaling form of the spin - spin correlation function ; the amplitude of staggered spin - spin correlation becomes an almost constant value within the length scale @xmath202 and decays exponentially for further distance than @xmath202 , that is , @xmath203 \quad { \rm for } \ \ r_{ij } > \xi , \label{eq : ssscaling}\end{aligned}\ ] ] where @xmath13 is a constant and @xmath204 is the distance between sites @xmath40 and @xmath45 along the chain measured in the cubic units . the staggered moments in eqs .
( [ eq : m_s^xy ] ) and ( [ eq : m_s^yz ] ) can be rewritten by using the spin - spin correlation .
for instance , eq .
( [ eq : m_s^xy ] ) is rewritten in the form @xmath205^{1/2 } \nonumber \\ & \sim & \bigg [ \frac{1}{\sqrt2 l } \int \big| \langle { \mbox{\boldmath $ s$}}_0 \cdot { \mbox{\boldmath $ s$}}_r
\rangle \big| dr \bigg]^{1/2 } , \label{eq : m_s^xy ss}\end{aligned}\ ] ] by introducing the continuum limit to evaluate the summation .
note that @xmath206 is the length of the chains in the @xmath207 system .
therefore , by assuming the scaling form in eq .
( [ eq : ssscaling ] ) , we obtain @xmath208^{1/2 } = \sqrt{c } \label{eq : scaling near}\ ] ] for the case of @xmath209 , and @xmath210^{1/2 } \nonumber \\ & = & \bigg [ \frac{1}{\sqrt2 l } \big\ { 2c\xi - ce\xi \exp\big(-\frac{\sqrt2 l}{\xi } \big ) \big\ } \bigg]^{1/2 } \nonumber \\ & \sim & \frac{2^{1/4 } \sqrt{c\xi}}{\sqrt{l } } \label{eq : m_s^xy ss2 } \label{eq : scaling far}\end{aligned}\ ] ] for the case of @xmath211 . the staggered moment in the @xmath164 chains in eq .
( [ eq : m_s^yz ] ) is also estimated in a similar manner . with considering the above arguments based on the scaling form ( [ eq : ssscaling ] )
, we can roughly estimate @xmath202 from the crossover of the scaling from @xmath197 [ eq . ( [ eq : scaling far ] ) ] to the constant [ eq .
( [ eq : scaling near ] ) ] ( + corrections ) in the system - size dependence in fig .
[ fig : m_s chain vs l^-0.5 ] .
that is , the linear dimension of the system size where the crossover occurs gives a rough estimate of @xmath202 .
for instance , at @xmath212 , @xmath193 deviates from the scaling of @xmath197 at @xmath213 , on the contrary , @xmath194 obeys the scaling down to @xmath214 .
this suggests that @xmath215 ( @xmath127 sites ) and @xmath216 ( @xmath217 sites ) .
[ note that the longest distance along the chain in the periodic @xmath207 system ( @xmath218 ) is @xmath219 . ]
the correlation length in the @xmath47 direction is about 3 times longer than in the @xmath164 and @xmath165 directions .
we note that the crossover length systematically shifts to a longer value as decreasing temperature , which suggests divergence of @xmath202 as @xmath220 .
[ @xmath202 is expected to diverge in all the directions with the same critical exponent although the divergence of @xmath221 is not clear in fig .
[ fig : m_s chain vs l^-0.5 ] ( b ) compared to @xmath222 . ]
therefore , we find a 1d anisotropy of the spin correlation : the magnitude of the moment is much enhanced and the correlation length is much longer in the @xmath47 chains than in the @xmath164 and @xmath165 chains . as predicted in the mean - field arguments , the staggered spin correlation develops dominantly in the @xmath47 chains due to the cooperation of the orbital ordering and the geometrical frustration .
recently , the neutron scattering experiment has been performed in both above and below @xmath18 .
@xcite a clear difference of the @xmath223 dependence of the inelastic neutron intensity has been found between the data above and below @xmath18 , and ascribed to the 1d spin anisotropy due to the orbital ordering .
the proposed pattern of the orbital order is the same as our result in sec .
[ sec : orbital ordering ] .
it has been recognized that the interplay between orbital and spin degrees of freedom often causes an anisotropic electronic state , which is observed in optical measurements .
for instance , in perovskite vanadium oxides , a strong 1d nature in the 3d lattice structure has been observed in the optical measurement , @xcite and theoretically ascribed to the anisotropic orbital exchange in the spin ordered state .
@xcite therefore , we expect that the 1d anisotropy found in the previous section also shows up in the optical response .
we consider the anisotropy of the optical response by calculating the spectral weight in the present spinel case .
the spectral weight is the total weight of the optical conductivity defined by @xmath224 where @xmath225 is the electron charge , @xmath226 is the planck s constant , @xmath227 is the diagonal element of the optical conductivity tensor , and @xmath228 or @xmath55 . here , we consider only the optical transfer within the @xmath8 @xmath20 electron bands and neglect that to the other bands such as the @xmath8 @xmath229 bands and the oxygen @xmath230 bands . in experiments ,
the integral of eq .
( [ eq : i_mu ] ) is calculated up to an appropriate energy cutoff to extract the spectral weight from the relevant @xmath20 levels . the spectral weight in eq .
( [ eq : i_mu ] ) is generally given by the kinetic energy in the @xmath20 hubbard model ( [ eq : h_multiorbital ] ) .
@xcite in the strong correlation limit , it is calculated by the spin and orbital exchange energy in the effective model ( [ eq : h_so ] ) .
@xcite we show in the following only the results of the calculations .
details of the calculations are explained in appendix b. in the calculation of the spectral weight for the present model , there are three points to be noticed .
one is that the original hubbard model ( [ eq : h_multiorbital ] ) contains not only the nearest - neighbor but also the third - neighbor hoppings .
these two types of hopping contribute differently to the spectral weight .
the second point is that the directions of the electron hoppings are different from the crystal axes .
electrons hop along the @xmath47 , @xmath164 , and @xmath165 chains which are canted by 45 degree from two crystal axes and perpendicular to the rest one . from this , for instance , @xmath231 is given by the kinetic energy both in the @xmath165 and @xmath164 chains but not in the @xmath47 chains .
the third point is the crystal symmetry .
since the system shows the tetragonal lattice distortion below @xmath109 , there holds a general relation in the form @xmath232 for @xmath143 .
we note that there remains weak anisotropy of eq .
( [ eq : ix , iy , iz ] ) even in the high - temperature phase above @xmath109 because of the broken cubic symmetry in @xmath51 . with taking account of these points , we can derive the following expressions @xmath233 , \label{eq : i_x } \\ i_y & = & \frac{-1}{2n_{\rm site } } \sum_{\zeta = xy , yz } \big [ \big\langle ( h_{\rm so}^{\rm nn})_{\zeta } \big\rangle + 4 \big\langle ( h_{\rm so}^{\rm 3rd})_{\zeta } \big\rangle \big ] , \label{eq : i_y } \\ i_z & = & \frac{-1}{2n_{\rm site } } \sum_{\zeta = yz , zx } \big [ \big\langle ( h_{\rm so}^{\rm nn})_{\zeta } \big\rangle + 4 \big\langle ( h_{\rm so}^{\rm 3rd})_{\zeta } \big\rangle \big ] , \label{eq : i_z } \end{aligned}\ ] ] where we set @xmath234 , and @xmath235 and @xmath236 represent the matrix elements of eqs .
( [ eq : h_nn ] ) and ( [ eq : h_3rd ] ) in the @xmath237 chains , respectively ( @xmath238 , @xmath165 , or @xmath47 ) . the derivation of eqs .
( [ eq : i_x])-([eq : i_z ] ) will be given in appendix b. in fig .
[ fig : spectral weight ] , we show the mc results for the spectral weight . the spectral weight becomes highly anisotropic in the tetragonal phase below @xmath109 with satisfying the relation ( [ eq : ix , iy , iz ] ) : the spectral weight in the @xmath55 direction , @xmath231 , is suddenly suppressed at @xmath109 , while those in the @xmath47 plane , @xmath239 and @xmath240 , are slightly enhanced there .
moreover , in the orbital ordered phase below @xmath109 , @xmath239 and @xmath240 increase monotonically with decreasing temperature , whereas @xmath231 shows a complicated temperature dependence although the dependence itself is small ; @xmath231 slightly decreases above @xmath115 , and it turns to increase below @xmath115 .
the increase below @xmath115 comes from the energy gain in the @xmath164 and @xmath165 directions by the development of the af spin order . , @xmath160 , and @xmath55 direction calculated by eqs .
( [ eq : i_x])-([eq : i_z ] ) at @xmath78 .
error bars are smaller than the symbol sizes .
the solid ( dashed ) line shows @xmath115 ( @xmath109 ) .
, width=264 ] the anisotropic electronic state indicated by our mc results can be observed in the optical measurements with applying the electric field in each direction .
it needs a clean surface in the single crystal .
unfortunately , it is difficult to grow a single crystal large enough thus far , and the experimental confirmation of our results remains for further study . here
, we discuss systematic changes with respect to the value of @xmath79 .
figures [ fig : j_3 dep of c ] , [ fig : j_3 dep of chi ] , and [ fig : j_3 dep of m_o and m_s ] show the specific heat in eq .
( [ eq : c ] ) , the uniform magnetic susceptibility in eq .
( [ eq : chi ] ) , and the orbital and magnetic moments in eqs .
( [ eq : m_o ] ) and ( [ eq : m_s ] ) , respectively , for several values of @xmath79 . the orbital and jt transition temperatures @xmath109 are indicated by the dashed lines in the figures , which are monitored by discontinuous changes of @xmath105 , @xmath242 , and @xmath138 as well as @xmath97 and @xmath149 ( not shown here ) .
the af transition temperatures @xmath115 are indicated by the downward arrows in the figures , which are determined by the singular peak of @xmath105 and a continuous development of @xmath162 as well as the diverging peak of @xmath174 and the crossing point of @xmath243 ( not shown here ) . in the case of @xmath244
, the system shows only the orbital and lattice transition , and there is no af ordering down to the lowest temperature . upon switching on @xmath80 ,
the magnetic transition appears at a finite temperature @xmath115 , below which the 3d af order exists . as shown in figs .
[ fig : j_3 dep of c]-[fig : j_3 dep of m_o and m_s ] , @xmath115 increases whereas @xmath109 slightly decreases as @xmath79 increases .
the increase of @xmath115 is easily understood because the af spin ordering is stabilized by the third - neighbor exchange interaction @xmath80 .
the reason for the decrease of @xmath109 is not so clear , but it might be due to the strong interplay between spin and orbital degrees of freedom as well as the frustration between @xmath245 and @xmath80 within the same chains . for @xmath246 , @xmath115 coincides with @xmath109 , and there is only one discontinuous transition where both orbital and af spin moments become finite discontinuously . , ( b ) @xmath247 , ( c ) @xmath248 , ( d ) @xmath249 , ( e ) @xmath250 , and ( f ) @xmath118 , respectively . the dashed lines ( the downward arrows ) indicate @xmath109 ( @xmath115 ) .
typical error bars are shown .
, width=302 ] , ( b ) @xmath247 , ( c ) @xmath248 , ( d ) @xmath249 , ( e ) @xmath250 , and ( f ) @xmath118 , respectively .
the dashed lines ( the downward arrows ) indicate @xmath109 ( @xmath115 ) .
error bars are smaller than the symbol sizes . , width=302 ] , ( b ) @xmath247 , ( c ) @xmath248 , ( d ) @xmath249 , ( e ) @xmath250 , and ( f ) @xmath118 , respectively .
the dashed lines ( the downward arrows ) indicate @xmath109 ( @xmath115 ) .
error bars are smaller than the symbol sizes .
, width=302 ] figure [ fig : phase diagram ] summarizes the phase diagram in the @xmath80-@xmath117 parameter space determined by the analysis of the results in figs .
[ fig : j_3 dep of c]-[fig : j_3 dep of m_o and m_s ] as well as other quantities such as @xmath97 , @xmath149 , @xmath174 , and @xmath243 .
the shaded area for @xmath187 shows the orbital ordered phase with the tetragonal jt distortion with a flattening of vo@xmath6 octahedra as shown in fig .
[ fig : orbital order ] ( a ) .
the 1d anisotropy of spin correlation found in sec .
[ sec:1d ] appears in this region .
the hatched area for @xmath200 is the af ordered phase concomitant with the orbital and lattice orders . )
determined by classical monte carlo calculations .
the shaded area @xmath187 shows the orbital and lattice ordered phase .
the hatched area below @xmath115 is the af ordered phase concomitant with the orbital and lattice orders .
the orbital and lattice transition at @xmath109 ( the dashed line ) is a first - order transition , while the af transition at @xmath115 ( the solid line ) is a second - order one .
the lines are guides for the eyes .
, width=264 ] from estimates of parameters in sec .
[ sec : parameters ] , we obtained @xmath75k and @xmath251k . from fig . [
fig : phase diagram ] , these estimates give @xmath252k and @xmath253k .
the experimental values are @xmath12k and @xmath14k in znv@xmath1o@xmath2 .
@xcite thus , two transitions at @xmath18 and @xmath15 in experiments are consistently understood by the orbital and lattice ordering transition at @xmath109 and the af ordering transition at @xmath115 , respectively .
the semiquantitative agreement of the transition temperatures between our mc results and the experimental values is satisfactory with considering the assumptions on the derivation of the model ( [ eq : h ] ) and on parameter estimates in sec .
[ sec : parameters ] .
in particular , we note that @xmath53 and @xmath56 in the jt part in eq .
( [ eq : h_jt ] ) , are parameters in our theory for which there has not been any experimental estimate to our knowledge as mentioned in sec .
[ sec : parameters ] .
the agreement in spite of these assumptions strongly suggests that our model ( [ eq : h ] ) captures essential physics in vanadium spinel oxides @xmath0v@xmath1o@xmath2 .
the orbital and lattice orderings at @xmath109 are considered to be caused by the cooperation between the intersite orbital interaction in eq .
( [ eq : h_so ] ) and the jt coupling in eq .
( [ eq : h_jt ] ) . in this section ,
we examine the role of the jt distortion in more detail .
we focus on the instability in the orbital sector and neglect spins momentarily .
that is , we here consider the effective orbital model which is derived from the model ( [ eq : h_so ] ) by assuming the spin paramagnetic state in the mean - field level as discussed in ref . .
we replace @xmath254 by @xmath255 in eq .
( [ eq : h_so ] ) , and obtain the effective orbital hamiltonian in the form @xmath256 where @xmath257 . for simplicity , we neglect small contributions from the @xmath80 terms in this section .
since @xmath258 , the hamiltonian is an af three - state clock model , which is the same as eq .
( 5 ) in ref . up to irrelevant constants .
the mean - field argument for the model ( [ eq : h_o ] ) predicts that a degeneracy remains partially in the tetrahedron unit .
@xcite there are totally @xmath259 different orbital states in the four - sites unit , and @xmath260 states among them are in the lowest energy state which have four antiferro - type and two ferro - type orbital bonds .
the @xmath260 degenerate states are categorized into two different types shown in fig . 2 ( a ) and ( b ) in ref .
[ the former corresponds to fig .
[ fig : orbital order ] ( b ) and the latter corresponds to figs .
[ fig : orbital order 2 ] ( a ) and ( b ) in the present paper . ] in the first type , two ferro - type bonds do not touch with each other , while they touch at one site in the second type .
thus , we expect that in the absence of the jt coupling , the orbital degeneracy remains partially and prevents the emergence of a particular ordered state . to confirm this prediction , we perform the monte carlo simulation for the effective orbital model ( [ eq : h_o ] ) . the result of the specific heat per site
is shown in fig .
[ fig : c and s ] ( a ) ( black symbols ) .
the data show a broad peak around @xmath261 without any singularity or any significant system - size dependence .
this indicates that the system does not show any phase transition , as predicted by the above mean - field picture . in fig .
[ fig : c and s ] ( b ) ( black symbols ) , we plot the temperature dependence of the entropy sum defined by @xmath262 the entropy sum appears to saturate at @xmath263 at high temperatures , which is largely suppressed from the expected value of @xmath264 for free three - state clock spins .
this indicates that there remains the entropy at zero temperature due to the degeneracy discussed above . ) ] for the effective orbital model ( [ eq : h_o ] ) .
black symbols show the results without the jt distortion .
gray symbols show the results in the presence of the tetragonal jt level splitting with @xmath265 in eq .
( [ eq : h_d ] ) . ,
width=264 ] one way to estimate the zero - temperature entropy @xmath266 is the so - called pauling s method .
@xcite in this method , the ground state degeneracy in the whole system is calculated simply multiplying the local degeneracy . in the present case ,
the total number of states is @xmath267 , and @xmath260 of @xmath268 states constitute the degenerate ground - state in each tetrahedron as mentioned above . with noting that there are @xmath269 tetrahedra in the system and assuming that the tetrahedra are independent with each other
, the zero - temperature entropy within the pauling s approximation is obtained as @xmath270 \nonumber \\ & = & \ln 3 - \frac12 \ln \big ( \frac{81}{30 } \big ) .
\end{aligned}\ ] ] thus , we obtain the entropy sum as @xmath271 the estimate is close to the saturated value in fig .
[ fig : c and s ] ( b ) . the tetragonal jt distortion with a flattening of the octahedra as shown in fig .
[ fig : orbital order ] ( a ) splits the degenerate energy levels , and lowers the energy of one orbital state of the first type , i.e. , the configuration in fig .
[ fig : orbital order ] ( b ) . in this case , all the tetrahedra are in the same orbital states .
that is , there we expect the orbital ordering and disappearance of the zero - temperature entropy .
the results for the case with the level splitting are also shown in fig .
[ fig : c and s ] ( gray symbols ) . here , for simplicity , we incorporate the level splitting by adding the term @xmath272 to eq .
( [ eq : h_o ] ) , which mimics the jt term in eq .
( [ eq : h_jt ] ) . as expected , the specific heat in fig .
[ fig : c and s ] ( a ) shows a singularity at @xmath273 , which indicates a phase transition .
figure [ fig : c and s ] ( b ) shows @xmath274 , which indicates that the phase transition reduces the remaining entropy by lifting the degeneracy .
the results in this section confirm the mean - field picture in ref .
, and explicitly show the importance of the cooperation between the intersite orbital interaction and the jt coupling . here
, we comment on the spatial symmetry of the orbital ordered state below @xmath109 .
our result in fig .
[ fig : orbital structure ] indicates that the orbital ordering breaks the mirror symmetry for the @xmath64 $ ] or @xmath275 $ ] plane .
the symmetry breaking will also appear in the lattice structure through the electron - phonon coupling which breaks the @xmath27 and @xmath29 symmetry although such coupling is not included in our model [ the tetragonal jt mode in eq .
( [ eq : h_jt ] ) does not split the @xmath27 and @xmath29 levels ] .
recently , the orbital ordered state with another symmetry has been proposed theoretically .
@xcite the theory is based on the assumption of the dominant role of the relativistic spin - orbit coupling .
the predicted orbital order consists of the ferro - type occupation of the @xmath31 and @xmath276 orbitals at every site .
this orbital order does not break the mirror symmetry .
the difference of the resultant orbital ordered state is important to consider the fundamental physics of the present @xmath20 electron system . in @xmath20 electron systems , in general , there is keen competition among different contributions whose energy scales are close to each other ; the spin and orbital exchange interactions , the jt coupling , and the relativistic spin - orbit coupling .
@xcite in our argument , the former two mechanisms play a primary role while the last one is not taken into account . on the contrary , in ref .
, the relativistic spin - orbit coupling is assumed to be dominant .
therefore , the determination of the orbital and lattice symmetry is relevant to clarify the primary factor in the keen competition .
experimental results on the lattice symmetry , however , are still controversial . in refs . and , the x - ray scattering results for polycrystal samples are consistent with the symmetry @xmath277 below @xmath18 , which suggests the persistence of the mirror symmetry at the low temperature phase . on the contrary , in recent synchrotron x - ray data for a single crystal sample ,
a new peak is found whose intensity is three orders of magnitude weaker than a typical main peak .
@xcite unfortunately , the new peak can not conclude whether the mirror symmetry is broken or not , but it suggests the different symmetry from the previous @xmath277 , i.e. , either @xmath278 or @xmath279 .
this controversy , probably coming from a small electron - phonon coupling in this @xmath20 system , reveals the necessary of more sophisticated experiments for larger single crystals .
such experiments are highly desired to settle the theoretical controversy .
another possible experiment to conclude the symmetry problem is to detect the orbital state directly .
this may be achieved , for instance , by the resonant x - ray scattering technique .
@xcite besides , another way to distinguish the orbital ordered states in our results and in ref .
is to examine the time reversal symmetry . the @xmath276 orbital order in ref
. breaks the time reversal symmetry even above the magnetic transition temperature .
a possible experiment is the detection of either circular dichroism or birefringence .
our classical monte carlo calculations have shown that with approaching zero temperature , the staggered moment increases and saturates at the maximum value , @xmath280 , when the third - neighbor coupling @xmath80 is finite . in physical units ,
this value should be multiplied by @xmath281 and corresponds to @xmath282 . here , @xmath283 is the bohr magneton and the @xmath284-factor is assumed to be the standard value @xmath285 .
this behavior was plotted in figs .
[ fig : spin ] ( a ) and [ fig : j_3 dep of m_o and m_s ] ( b)-(f ) . on the other hand ,
recent experiments show @xmath286 at low temperatures , only less than a half of this classical value . @xcite it is well known that quantum fluctuations due to magnon excitations reduce the amplitude of spontaneous moment in antiferromagnets .
we expect that this effect is particularly important in frustrated magnets like pyrochlore systems , since frustration generally reduces the energy scale of magnon excitations and correspondingly enhances quantum fluctuations of the staggered moment . here
, we examine at @xmath106 the effect of quantum fluctuations of spin degree of freedom on the reduction of staggered moment for the model ( [ eq : h ] ) , by using the linear spin wave theory . for this purpose
, we assume the perfect orbital order with the ordering pattern in fig .
[ fig : orbital structure ] , and substitute the density operators in eq .
( [ eq : h_so ] ) by their expectation values 0 or 1 .
that is , we here consider the effective quantum spin model on the pyrochlore lattice in the form @xmath287 where the first summation should be taken over nearest - neighbor pairs , while the second one be over third - neighbor pairs coupled by @xmath34-bond . under the orbital ordering in fig .
[ fig : orbital structure ] , exchange coupling constants are ferromagnetic for nearest - neighbor spin pairs on the @xmath164 and @xmath165 chains , @xmath288 , and antiferromagnetic for those on the @xmath47 chains , @xmath289 , while the third - neighbor couplings are also antiferromagnetic , @xmath290 .
these parameters were defined in eqs .
( [ eq : j])-([eq : eta ] ) .
it is important that @xmath291 , since @xmath292 and @xmath293 . roughly speaking ,
the small control parameter in the spin wave theory is 1/[@xmath294 ( number of neighbors ) ] . in the present case ,
@xmath295=1 and @xmath55=6 , leading to a control parameter small enough , and therefore we may expect quite good estimate from the spin wave theory . for the magnetic order determined by the mean field arguments in sec .
[ sec : spin - orbital model ] and ref .
, we have calculated the reduction of staggered moment at @xmath106 by using the linear spin wave theory .
the magnetic structure is shown in fig .
[ fig : spin order ] and its unit cell contains 8 spins .
details of the calculations will be reported elsewhere and here we show only the results for the staggered moment . when the third - neighbor exchange couplings are zero ( @xmath296 ) , magnons have zero energy ( zero modes ) on the planes @xmath297 in the brillouin zone .
within the linear spin wave theory , magnons are treated as noninteracting to each other , and these zero modes are excited freely without any energy cost .
this leads to the logarithmic divergence of the reduction of moment , @xmath298 .
it is noted that in the pyrochlore spin system in which all the nearest - neighbor couplings are antiferromagnetic with same amplitude , a half of magnon excitations are zero modes throughout the whole brillouin zone , and this results in @xmath299 , as a manifestation of strong geometrical frustration .
@xcite once the third - neighbor exchange couplings @xmath300 are switched on , zero modes acquire a positive energy and the reduction of moment becomes finite .
figure [ fig : deltas ] shows the results of the reduction of moment @xmath301 as a function of @xmath300 for @xmath302 and @xmath303 , corresponding to @xmath304 at which the mc calculations have been performed .
the amplitude of staggered moment is reduced to @xmath305 . in the case of the vanadium spinel oxides with @xmath9 , @xmath306 for @xmath307 , which indicates that the antiferromagnetic order shown in fig . [
fig : spin order ] is stable down to very small third - neighbor exchange couplings .
for example , at a reasonable value , @xmath308 , the amplitude of staggered moment is @xmath309 .
with using @xmath285 , this corresponds to @xmath310 .
hence , the reduction is significantly large in the present spin - orbital - lattice coupled system , and particularly the staggered moment rapidly decreases as @xmath300 in the realistic small-@xmath300 region .
we consider that the estimate of @xmath162 is satisfactory and can explain the experimental value by carefully tuning the model parameters . .
the staggered moment is also plotted ( filled symbols ) , which is given by @xmath305 and by assuming @xmath285 .
the lines are guides for the eyes .
, width=264 ] we also note that there may be a small contribution from quantum fluctuations in the orbital degree of freedom . in our spin - orbital - lattice coupled model ( [ eq : h ] ) , this contribution is not taken into account since we consider only @xmath34 bond for hopping integrals .
other hopping integrals lead to small orbital exchange interactions , causing small quantum fluctuations of orbitals .
this will slightly reduce the orbital sublattice moment and may lead to a further reduction of the staggered spin moment .
we have shown that the mc results for the uniform magnetic susceptibility in sec .
[ sec : chi ] qualitatively explain the experimental data in the zero - field - cool ( zfc ) measurement . in experiments
, there has been reported that the zfc and field - cool ( fc ) data show a large difference at low temperatures .
@xcite the difference between the zfc and fc data emerges at a higher temperature than the structural transition temperature @xmath18 , and develops on cooling .
one of the mysteries is that the difference remains finite and large even in the af spin ordered phase below @xmath15 : the fc value becomes nearly twice of the zfc one at the lowest temperature .
this strongly suggests that the difference can not be explained as a simple spin - glass phenomenon .
a key observation is that the starting temperature of the difference between the zfc and fc data as well as the magnitude of the difference appears to depend on samples . in ref .
, the difference emerges at @xmath311k , and becomes @xmath312emu / v - mol at @xmath18 . on the other hand , in ref . , the difference starts at @xmath313k , and becomes @xmath314emu / v - mol at @xmath18 .
we also note that the temperature dependences of the zfc data show a sample dependence , especially near two transition temperatures .
these sample dependences imply the importance of quenched disorder .
effects of disorder in the present spin - orbital - lattice coupled system are interesting and open problems . in our mc calculations
, it is difficult to obtain the fc results because the system shows a discontinuous transition at @xmath109 : there appears a huge hysteresis when we change the temperature successively by using the mc sample in the previous run as the initial state in the next run . on cooling ,
the system remains in the high - temperature para state well below the true transition temperature @xmath109 as a metastable state because of an exponentially large energy barrier in the first - order transition .
one way to avoid this hysteresis is to apply more sophisticated mc method such as the multicanonical technique .
@xcite this interesting problem is left for further study .
we have shown that our spin - orbital - lattice coupled model ( [ eq : h ] ) exhibits two transitions which well agree with experimental results in @xmath0v@xmath1o@xmath2 . here , we discuss a difference among the compounds with different @xmath0 cations observed in experiments . among @xmath0v@xmath1o@xmath2 with @xmath0=zn ,
mg , or cd , cdv@xmath1o@xmath2 shows a different behavior near @xmath18 from others .
@xcite as temperature decreases , the magnetic susceptibility for the compounds with @xmath0=zn and mg shows a sudden drop at @xmath18 as in our numerical result in fig .
[ fig : chi ] , whereas that for the cd compound shows a sharp increase .
moreover , the transition temperature @xmath18 is substantially higher in the cd case than in the zn and mg cases ; @xmath315k for cd while @xmath316k and @xmath317k for zn and mg , respectively . on the other hand ,
the af magnetic transition temperature @xmath15 is almost the same among three compounds ; @xmath318 , and @xmath319k for zn , mg , and cd , respectively .
@xcite our results show that the transition temperature @xmath109 , which corresponds to @xmath18 , is determined by the energy balance between the intersite orbital interaction in eq .
( [ eq : h_so ] ) and the elastic energy gain in eq .
( [ eq : h_jt ] ) .
the @xmath0-site cations are nonmagnetic and locate at the tetrahedral sites which are separated from the pyrochlore network of v cations , and the change of the @xmath0 cation may affect the lattice structure since @xmath0o@xmath2 tetrahedra share oxygen ions with vo@xmath6 octahedra . thereby , we may consider that the cd cation , which has relatively large ionic radius , may modify the lattice structure and have an influence on the jt energy gain as well as the form and magnitude of the orbital interaction . actually , although the symmetry is the same for the three compounds , the so - called @xmath320 parameter of the spinel structure is considerably larger for cdv@xmath1o@xmath2 than for zn and mg compounds .
the @xmath320 parameter represents the magnitude of the trigonal distortion .
the absence of the trigonal distortion gives @xmath321 by definition , and a larger @xmath320 denotes a larger trigonal distortion .
for the zn and mg cases , @xmath322 and @xmath323 , respectively , whereas @xmath320 becomes 0.394 for the cd compound : the trigonal distortion is substantially larger in the cd case than in the zn and mg cases .
this suggests that the peculiar behavior in cdv@xmath1o@xmath2 may be ascribed to effects of the trigonal distortion .
the agreement between the experimental data in zn and mg compounds and our numerical results , which do not include effects of the trigonal distortion , indicates that the small trigonal distortion does not play an important role in these two compounds . to understand the behavior of the cd compound , i.e. , the sharp increase in the magnetic susceptibility as well as the higher @xmath18 , it is interesting to examine the effect of the trigonal distortion in our theoretical framework .
in addition to the contribution from the electron - phonon coupling to the trigonal distortion , we may have to include more complicated hopping integrals not only the @xmath34-bond type but also , for instance , @xmath324-bond type or the second - neighbor hoppings , since the trigonal distortion affects the network of vo@xmath6 octahedra : these additional hoppings modify the spin and orbital intersite interactions in the derived effective model in eq .
( [ eq : h_so ] ) .
the complicated effects of this trigonal distortion are left for further study .
in the present work , we have investigated the microscopic mechanism of two transitions in vanadium spinel oxides @xmath0v@xmath1o@xmath2 with nonmagnetic divalent @xmath0 cations such as zn , mg , and cd .
we have focused on the role of the @xmath20 orbital degree of freedom as well as spin in these strongly correlated electron systems on the geometrically frustrated lattice structure , i.e. , the pyrochlore lattice consists of the magnetic v cations . we have derived the effective spin - orbital - lattice coupled model in the strong correlation limit from the multiorbital hubbard model with explicitly taking account of the @xmath20 orbital degeneracy .
the effective model describes the interplay between orbital and spin , and reveals the contrasting form of the intersite interactions in two degrees of freedom ; the heisenberg type for the spin part and the three - state clock type for the orbital part .
the anisotropy in the orbital interaction originates from the dominant role of the @xmath34-bond hopping integrals in the edge - sharing configuration of vo@xmath6 octahedra . the jahn - teller coupling with the tetragonal lattice distortion is also included .
thermodynamic properties of the effective model have been investigated by the monte carlo simulation , which is a classical one to avoid the negative sign problem due to the geometrical frustration .
quantum corrections are examined by the spin wave approximation .
main results are summarized in the following .
our effective spin - orbital - lattice coupled model exhibits two transitions at low temperatures .
as temperature decreases , first , the orbital ordering transition occurs assisted by the tetragonal jahn - teller distortion .
this is a first - order transition .
successively , at a lower temperature , the antiferromagnetic spin order sets in .
this transition is second order . for realistic parameter values ,
our numerical results agree with experimental data semiquantitatively .
the estimates of two transition temperatures are @xmath252k and @xmath253k , while the experimental values are @xmath325k and @xmath326k .
the changes of the entropy at the transitions are comparable to the experimental values : in particular , the small change at @xmath115 compared to the considerable change at @xmath109 is well reproduced .
the magnetic ordering structure in the low temperature phase is completely consistent with the neutron scattering results .
the magnitude of the staggered moment at @xmath106 is largely reduced by quantum fluctuations , and the estimate by the spin wave theory reasonably agrees with the experimental data .
the temperature dependence of the uniform magnetic susceptibility is similar to the experimental data .
the agreement strongly indicates that our effective model captures the essential physics of the vanadium spinel compounds @xmath0v@xmath1o@xmath2 .
our results give an understanding of the mechanism of the two transitions in the vanadium spinel oxides .
the present numerical study has confirmed the mean - field scenario in our previous publication , @xcite and moreover , given more detailed and quantitative information . the first transition with orbital and lattice orderings is induced by the intersite orbital interaction which is three - state clock type and spatially anisotropic depending on both the bond direction and the orbital states in two sites .
the anisotropy lifts the degeneracy due to the geometrical frustration inherent to the pyrochlore lattice .
the tetragonal jahn - teller coupling also plays an important role to stabilize the particular orbital ordering pattern .
the obtained orbital ordering structure is the antiferro type which consists of the alternative stacking of two ferro - type @xmath16 planes ; ( @xmath327 ) orbitals are occupied in one plane , and ( @xmath328 ) orbitals are occupied in the other .
once the orbital ordering takes place , the orbital state affects the spin exchange interactions through the spin - orbital interplay , and reduces the magnetic frustration partially .
as a consequence , the antiferromagnetic spin correlation develops mainly in the one - dimensional chains in the @xmath16 planes .
we found typically about three - times longer correlation length in the @xmath47 direction than in the @xmath164 and @xmath165 directions . at the second transition ,
these one - dimensional chains are ordered by the third - neighbor exchange interaction to form the three - dimensional antiferromagnetic spin order .
it was found that thermal fluctuations stabilize the collinear state by the order - by - disorder type mechanism .
there still remain several open problems on this topic as discussed in sec .
[ sec : discussions ] ; the controversy on the symmetry of the orbital ordered state , the large difference of the zero - field - cool and field - cool data of the magnetic susceptibility , and the @xmath0-site dependence , which is probably related to the trigonal distortion .
they need further investigations from both experimental and theoretical viewpoints . besides them
, another important problem is the carrier doping effect on the mott insulating materials , such as the li doping in zn@xmath329li@xmath330v@xmath1o@xmath2 . in experiments , it is known that the doping rapidly destroys the ordered states at @xmath331 , and replaces them by a glassy state .
@xcite finally , at @xmath332 , the system becomes metallic and shows a heavy - fermion like behavior .
@xcite we believe that our present results give a good starting point to study the interesting doping effects .
this work was supported by a grant - in - aid and naregi nanoscience project from the ministry of education , science , sports , and culture . a part of the work was accomplished during y. m. was staying at the yukawa institute of theoretical physics , kyoto university , with the support from the 21st century for center of excellence program , ` center for diversity and universality in physics ' .
in this appendix , we examine the conditions for @xmath53 and @xmath56 in the hamiltonian ( [ eq : h_jt ] ) to stabilize the orbital order in fig . [
fig : orbital order ] ( b ) accompanied by the tetragonal jt distortion with a flattening of vo@xmath6 octahedra as shown in fig .
[ fig : orbital order ] ( a ) at low temperatures .
we employ the mean - field type argument to discuss the instability at high temperatures as in ref . .
here , we consider only the nearest - neighbor interactions and neglect small contributions from @xmath62 . with considering a spin disordered state and replacing @xmath254 by @xmath255 in eq .
( [ eq : h ] ) , we obtain the effective hamiltonian for the orbital and jt parts as @xmath333 where @xmath257 .
the first term is equivalent to eq .
( [ eq : h_o ] ) and to eq .
( 5 ) in ref . up to irrelevant constants .
we consider the orbital state and the jt distortion in the tetrahedron unit shown in fig .
[ fig : orbital order ] ( b ) . for an orbital configuration at the four sites ,
the expectation value of the jt part @xmath51 for one tetrahedron is written in the following quadratic form @xmath334 where @xmath335 is a four - dimensional vector which denotes the amplitude of the jt distortion in each site as @xmath336 , and @xmath337 is the @xmath338 matrix whose matrix elements are given by @xmath339 for @xmath340 and @xmath341 for @xmath342 .
the vector @xmath343 describes the electron - phonon coupling part and depends on the orbital state ; for instance , @xmath344 for the orbital occupation in fig .
[ fig : orbital order ] ( b ) .
here , we consider the physical situation in which the system does not show any spontaneous lattice distortion without the coupling to electrons of the second and third terms in eq .
( [ eq : e_jt ] ) .
this is satisfied when the matrix @xmath337 is positive definite .
three of the eigenvalues of @xmath337 are @xmath345 and the last one is @xmath346 , and therefore , by noting that we consider only positive @xmath56 as mentioned in sec . [
sec : spin - orbital model ] , the condition is @xmath347 the quadratic form of @xmath348 in eq .
( [ eq : e_jt ] ) can be transformed to the following form @xmath349 where @xmath350 , @xmath351 , and @xmath352 is well - defined since @xmath337 is a positive - definite matrix under the condition of eq .
( [ eq : cond ] ) .
therefore , we find that the jt energy takes its minimum value @xmath353 for the amplitude of the distortions @xmath354 for instance , for the configuration in fig .
[ fig : orbital order ] ( b ) , we obtain @xmath355 for @xmath356 with @xmath357 . all the different orbital configurations with @xmath358 or @xmath125 are classified by the energy value @xmath348 into five groups , which are described by the vector @xmath343 as @xmath359 where arbitrary permutations of components in each vector give the same energy .
typical orbital configurations are shown in fig .
[ fig : orbital order 2 ] .
( the configuration for @xmath360 is shown in fig .
[ fig : orbital order ] ( b ) . ) on the basis of the above consideration on the jt energy , we calculate the expectation value of the hamiltonian ( [ eq : h_orb - jt ] ) , for the five types of orbital configurations . with considering the orbital interaction energy from the first term in eq .
( [ eq : h_orb - jt ] ) , the minimized values of the energy per tetrahedron are obtained as @xmath361 note that there are twelve bonds per one tetrahedron . by comparing @xmath362 with other four values @xmath363 ( @xmath364 )
, we obtain the stability conditions for the orbital configuration of fig .
[ fig : orbital order ] ( b ) .
first , the conditions @xmath365 and @xmath366 are satisfied when @xmath367 in this range , the condition @xmath368 leads to @xmath369 and the condition @xmath370 leads to @xmath371 the shaded area in fig .
[ fig : conditions ] shows the parameter region where all the conditions of eqs .
( [ eq : condition1])-([eq : condition3 ] ) are satisfied .
we performed several mc runs to confirm the present mean - field type analysis . in the mc calculations in sec .
[ sec : results ] , we take @xmath84 and @xmath85 as a typical set in this region . )
-([eq : condition3 ] ) is shown by the shaded area .
the dashed and solid lines denote the conditions of eqs .
( [ eq : condition2 ] ) and ( [ eq : condition3 ] ) , respectively . , width=245 ]
in this appendix , we derive the expressions for the spectral weights in eqs .
( [ eq : i_x])-([eq : i_z ] ) for our effective model ( [ eq : h ] ) . the spectral weight , which is the total weight of the optical conductivity , is given by the kinetic energy of the system ( @xmath372-sum rule ) .
@xcite we may safely extend the formula for the single - band hubbard model ; @xmath373 to our multiorbital case .
( we set @xmath234 , and @xmath374 is the number of unit cells . ) here , @xmath375 describes the kinetic energy term of the multiorbital hubbard hamiltonian as @xmath376 which is the fourier transform of the first term of eq .
( [ eq : h_multiorbital ] ) . here ,
@xmath377 are the atom indices in the primitive unit - cell containing four v atoms , whose relative positions are labelled by @xmath378 , @xmath379 , @xmath380 , @xmath381 , respectively . since the pyrochlore lattice consists of three different chains in the @xmath47 , @xmath164 , and @xmath165 directions , and we here consider only the @xmath34-bond hopping integrals which are finite only along the chains , @xmath375 is given by the summation of the contributions from each chain as @xmath382 , \label{eq : k}\ ] ] where @xmath383 ( @xmath164 ) , @xmath28 ( @xmath165 ) , @xmath30 ( @xmath47 ) denote the three different types of the chains , and @xmath384 are the nearest - neighbor and the third - neighbor hopping integrals along the @xmath237 chain , respectively . in the hamiltonian ( [ eq : h_multiorbital ] ) , @xmath385 and @xmath386 , but they are deliberately denoted by different parameters for later use .
each term in eq .
( [ eq : k ] ) is defined by @xmath387 \nonumber \\ & & \quad + 2 \cos({\mbox{\boldmath $ k$ } } \cdot { \mbox{\boldmath $ \delta$}}_{23 } ) [ c_{21}^\dagger({\mbox{\boldmath $ k$ } } ) c_{31}({\mbox{\boldmath $ k$ } } ) + c_{31}^\dagger({\mbox{\boldmath $ k$ } } ) c_{21}({\mbox{\boldmath $ k$ } } ) ] , \\ & & \kappa_{1}^{\rm 3rd } ( { \mbox{\boldmath $ k$ } } ) = 2 \cos(2{\mbox{\boldmath $ k$ } } \cdot { \mbox{\boldmath $ \delta$}}_{01 } ) [ c_{01}^\dagger({\mbox{\boldmath $ k$ } } ) c_{01}({\mbox{\boldmath $ k$ } } ) + c_{11}^\dagger({\mbox{\boldmath $ k$ } } ) c_{11}({\mbox{\boldmath $ k$ } } ) ] \nonumber \\ & & \quad + 2 \cos(2{\mbox{\boldmath $ k$ } } \cdot { \mbox{\boldmath $ \delta$}}_{23 } ) [ c_{21}^\dagger({\mbox{\boldmath $ k$ } } ) c_{21}({\mbox{\boldmath $ k$ } } ) + c_{31}^\dagger({\mbox{\boldmath $ k$ } } ) c_{31}({\mbox{\boldmath $ k$ } } ) ] , \end{aligned}\ ] ] and so on , where @xmath388 .
let us now evaluate the derivatives of @xmath375 in eq .
( [ eq : i ] ) . by using eq .
( [ eq : k ] ) , it is straightforward to obtain , for instance , the derivative in terms of @xmath389 in the form @xmath390 \nonumber \\ & + & \frac14 \big [ v_2 \kappa_2^{\rm 3rd}({\mbox{\boldmath $ k$ } } ) + v_3 \kappa_3^{\rm 3rd}({\mbox{\boldmath $ k$ } } ) \big ] .
\end{aligned}\ ] ] the derivatives in terms of @xmath391 and @xmath392 are obtained in a similar manner .
thus , with noting @xmath393 , the spectral weights for the multiorbital hubbard model ( [ eq : h_multiorbital ] ) are obtained in the forms @xmath394 , \label{eq : i_x hub } \\ i_y & = & -\frac{1}{4n_{\rm site } } \big [ \big\ { \langle ( h_{\rm k}^{\rm nn})_{xy } \rangle + \langle ( h_{\rm k}^{\rm nn})_{yz } \rangle \big\ } \nonumber \\ & & \quad \quad + 4 \big\ { \langle ( h_{\rm k}^{\rm 3rd})_{xy } \rangle + \langle ( h_{\rm k}^{\rm 3rd})_{yz } \rangle \big\ } \big ] , \label{eq : i_y hub } \\ i_z & = & -\frac{1}{4n_{\rm site } } \big [ \big\ { \langle ( h_{\rm k}^{\rm nn})_{yz } \rangle + \langle ( h_{\rm k}^{\rm nn})_{zx } \rangle \big\ } \nonumber \\ & & \quad \quad + 4 \big\ { \langle ( h_{\rm k}^{\rm 3rd})_{yz } \rangle + \langle ( h_{\rm k}^{\rm 3rd})_{zx } \rangle \big\ } \big ] , \label{eq : i_z hub } \end{aligned}\ ] ] where @xmath395 and @xmath396 are the nearest - neighbor and the third - neighbor matrix elements of the kinetic term in eq . ( [ eq : h_multiorbital ] ) along the @xmath237 chain , respectively . in the strong correlation limit
, the kinetic energy corresponds to the spin and orbital exchange energy as discussed below .
the expectation values of the kinetic terms in eqs .
( [ eq : i_x hub])-([eq : i_z hub ] ) can be represented as coupling - constant derivatives ; @xmath397 here , we have used the hellman - feynman theorem to exchange the order of derivative and expectation value . in the strong correlation limit , we can replace the derivatives in terms of @xmath398 and @xmath399 of @xmath400 by the derivatives in terms of the exchange @xmath245 of the expectation values of the effective spin - orbital hamiltonian ( [ eq : h_so ] ) as @xmath401 where @xmath402 and @xmath403 are the nearest - neighbor and the third - neighbor exchange coupling constants along the @xmath237 direction , respectively ; and @xmath235 and @xmath236 are the nearest - neighbor and the third - neighbor matrix elements of eqs .
( [ eq : h_nn ] ) and ( [ eq : h_3rd ] ) along the @xmath237 chain , respectively . in the model ( [ eq : h_so ] ) , @xmath404 and @xmath405 , and they are independent of the direction @xmath237 . by substituting all the expectation values in eqs .
( [ eq : i_x hub])-([eq : i_z hub ] ) by eqs .
( [ eq : h_hub to h_so 1 ] ) , ( [ eq : h_hub to h_so 2 ] ) , and the similar expressions , we obtain the formulas of the spectral weights for the effective spin - orbital coupled model in the strong correlation limit as given in eqs .
( [ eq : i_x])-([eq : i_z ] ) .
s. kondo , d.c .
johnston , c.a .
swenson , f. borsa , a.v .
mahajan , l.l .
miller , t. gu , a.i .
goldman , m.b .
maple , d.a .
gajewski , e.j .
freeman , n.r .
dilley , r.p .
dickey , j. merrin , k. kojima , g.m .
luke , y.j .
uemura , o. chmaissem , and j.d .
jorgensen , phys .
lett * 78 * , 3729 ( 1997 ) ; s. kondo , d.c .
johnston , and l.l .
miller , phys .
b * 59 * , 2609 ( 1999 ) .
note that the definition of @xmath80 is different from that in ref . for convenience .
@xmath406 in the present work corresponds to @xmath80 in ref . .
note also that eqs .
( [ eq : a ] ) and ( [ eq : b ] ) are slightly modified from those in ref .
because of a minor mistake in the perturbation calculation , however , these minor changes do not alter the mean - field results in ref . . in the orbital ordered state below @xmath109
, there are two energy minima with equal energy ; one is for the orbital occupations in figs .
[ fig : orbital ] ( b)-(e ) and the other is for the occupations with exchanging @xmath407 and @xmath408 . in the present mc calculations ,
one of two configurations is selected depending on the initial configuration , and the mc sampling is performed around the selected energy minimum , since the energy barrier between the two minima is exponentially large in the first order transition .
y. murakami , h. kawada , h. kawata , m. tanaka , t. arima , y. moritomo , and y. tokura , phys .
lett . * 80 * , 1932 ( 1998 ) ; y. murakami , j.p .
hill , d. gibbs , m. blume , i. koyama , m. tanaka , h. kawata , t. arima , y. tokura , k. hirota , and y. endoh , phys .
lett . * 81 * , 582 ( 1998 ) . | we present our theoretical and numerical results on thermodynamic properties and the microscopic mechanism of two successive transitions in vanadium spinel oxides @xmath0v@xmath1o@xmath2 ( @xmath0=zn , mg , or cd ) obtained by monte carlo calculations of an effective spin - orbital - lattice model in the strong correlation limit .
geometrical frustration in the pyrochlore lattice structure of v cations suppresses development of spin and orbital correlations , however , we find that the model exhibits two transitions at low temperatures .
first , a discontinuous transition occurs with an orbital ordering assisted by the tetragonal jahn - teller distortion .
the orbital order reduces the frustration in spin exchange interactions , and induces antiferromagnetic correlations in one - dimensional chains lying in the perpendicular planes to the tetragonal distortion . secondly , at a lower temperature , a three - dimensional antiferromagnetic order sets in continuously , which is stabilized by the third - neighbor interaction among the one - dimensional antiferromagnetic chains .
thermal fluctuations are crucial to stabilize the collinear magnetic state by the order - by - disorder mechanism .
the results well reproduce the experimental data such as transition temperatures , temperature dependence of the magnetic susceptibility , changes of the entropy at the transitions , and the magnetic ordering structure at low temperatures .
quantum fluctuation effect is also examined by the linear spin wave theory at zero temperature .
the staggered moment in the ground state is found to be considerably reduced from saturated value , and reasonably agrees with the experimental data . |
progress in cosmic ray ( cr ) physics@xcite has been hampered for long time by the deflection of charged cosmic rays in magnetic fields , preventing the identification of individual sources .
this problem could be solved by using the neutral messengers that should be produced as secondaries in hadronic cr interactions close to the source . however , the secondary photons generated are difficult to disentangle from photons produced by synchrotron radiation or inverse compton scattering of electrons .
moreover , high energy photons are strongly absorbed both in the source and propagating over extragalactic distances . by contrast , the extremely large mean free path of neutrinos together with the relatively poor angular resolution of neutrino telescopes and the small expected event numbers makes the identification of extragalactic sources challenging using only the neutrino signal .
performing neutrino astronomy beyond the establishment of a diffuse neutrino background requires therefore most likely additional input , either timing or angular information from high energy photon or cr experiments . the recently announced evidence@xcite for a correlation of the arrival directions of uhecrs observed by the pierre auger observatory ( pao ) with active galactic nuclei ( agn ) may provide a first test case for successful `` multi - messenger astronomy . ''
in particular , ref
. finds two events within the search bin of @xmath3 around the nearest active galaxy , centaurus a ( cen a ) .
this fr i radio galaxy is located close to the supergalactic plane at a distance of about 4 mpc ( see ref . for more details ) . at present this correlation ,
not confirmed by the hires experiment@xcite , has only @xmath4 c.l . and
other source types that follow the large - scale structure of matter would also result in an excess of events along the supergalactic plane . in ref .
, we studied therefore the possibility to observe cen a using high - energy photons and neutrinos . taking the correlation signal at face value
, we used the pao results as normalization of the cr flux and calculated the flux of accompanying secondary photons and neutrinos expected from hadronic interactions in the source . since both the fermi lat collaboration @xcite and the h.e.s.s .
collaboration @xcite reported recently the discovery of @xmath5-ray emission from cen a , we have now the opportunity to check our predictions against these observations .
we calculated the flux of high energy cosmic rays and of accompanying secondary photons and neutrinos expected from cen a for two different scenarios : acceleration close to the core , either in accretion shocks or regular electromagnetic fields , and acceleration in the radio jet .
in the first case uv photons are the most important scattering targets .
we modeled the primary photon field around the agn core guided by the simplest possible theoretical model@xcite , namely the thermal emission from a geometrically thin , optically thick keplerian accretion disc .
the interaction depth for photo - hadron interactions can reach @xmath6 few . according to the observational data we found that pp interactions of uhe protons with the gas provide the main source of cr interactions in the second scenario . in this case ,
moreover , diffusion in the turbulent magnetic fields will increase the interaction depth at low and intermediate energies .
we considered also three spectra @xmath7 of the injected protons : power - laws with @xmath8 and @xmath9 , and a broken power - law with @xmath10 for @xmath11ev .
we based our calculations on several simplifying assumptions like the use of an one - dimensional geometry and the omission of the acceleration process .
in particular , we only postulated that acceleration to @xmath12ev is possible in the environment of cen a , without demonstrating it for a concrete model .
finally , hadronic interactions are simulated with an extension of the monte carlo code described in ref . .
figure [ spec1 ] displays the particle fluxes predicted@xcite from cen a as function of the energy , assuming that the two events observed by pao around cen a indeed originate from this agn .
the case of acceleration close to the core is shown on the left , while the case of acceleration in the jet is shown on the right . from the top to the bottom , spectra are displayed for a broken power - law , @xmath9 , and @xmath8 .
in addition to the injected proton flux ( black solid line ) , we show the flux of protons ( black dashed ) , photons ( blue solid ) and neutrinos ( red solid ) arriving on earth .
note that the cutoff in the neutrino and proton spectra below 100gev is artificial , since we neglect neutrinos and protons with lower energies in our simulation . in the core model
the final proton flux is reduced by photon - proton interactions by a factor @xmath13 above the threshold energy @xmath14ev ( left ) , while diffusion in the jet increases the interaction depth for lower energies ( right ) , resulting in the effective production of secondaries . since the cr spectra are normalized to the integral uhecr flux above @xmath15ev , steeper spectra result in larger secondary fluxes at low energies .
while we could show earlier@xcite these fluxes only together with an upper limit from h.e.s.s . and the estimated sensitivity of fermi for point sources , we have updated now these figures with the recently published results from fermi@xcite and h.e.s.s .
+ citeaharonian:2009xn . remarkably , the photon flux in the fermi and the h.e.s.s .
energy range has approximately the same power - law exponent ( @xmath16 and 2.7 ) , but the latter requires a larger normalization constant .
such a behavior is expected , if a new component , e.g. of hadronic origin , sets in above 100 gev , while at lower energies photons of electromagnetic origin dominate the spectrum . for a more detailed test of this hypothesis in the future , the differential energy spectrum at the high - energy end of the fermi spectrum will be most useful . the most important consequence of the recent h.e.s.s
. results is to significantly disfavor the jet scenario . in spite of the uncertainties in the normalization , the almost flat spectrum predicted in this model
is in contradiction with the shape of the @xmath5-ray spectrum observed by h.e.s.s .
moreover , the angular extension of the photon flux observed by h.e.s.s . is consistent with a point source at the agn core and excludes thereby the radio lobes as sources . on the contrary , the shape of the @xmath5-ray spectrum observed by h.e.s.s .
agrees very well with the slope expected in the core model .
the fermi measurements restrict additionally the source model , excluding the broken - power law case that leads to an excessive photon fluxes in the gev range . in summary , a primary proton energy @xmath0 with @xmath1 or smaller in the case of acceleration close to the core
is consistent with the h.e.s.s .
and fermi observations , assuming that in the gev range the photon flux has dominantly an electromagnetic origin .
finally , we comment on the neutrino fluxes to be expected from cen a. calculating the expected event number in a neutrino telescope requires a definite choice of the experiment .
centaurus a is from the location of icecube only visible from above , and thus the background of atmospheric muons allows only the use of contained events that carry essentially no directional information . by contrast , a neutrino telescope in the mediterranean could make use of the muon signal and the directional information .
the resulting event numbers both for cascade and shower event number per year observation time are summarized in table [ tablei ] . for a discussion on the connection between cen
a as source of uhecrs and the associated diffuse neutrino flux see ref . .
in summary , the combination of data from pao and h.e.s.s . is consistent with a scenario where the uhecrs observed by pao are protons accelerated near the core .
the vhe @xmath5-rays detected by h.e.s.s .
would then have a hadronic origin , namely , the interaction of the accelerated protons with the uv photons surrounding the core .
this model is expected to be tested in the near future as the currently running pao , h.e.s.s .
and fermi experiments increase their statistics .
s.o . acknowledges a marie curie ief fellowship from the european community , r.t .
support from the deutsche forschungsgemeinschaft within the sfb 676 . 0 v. s. berezinsky , s. v. bulanov , v. a. dogiel , v. l. ginzburg , v. s. ptuskin , _ astrophysics of cosmic rays _ , ( north - holland , amsterdam 1990 ) ; m. nagano and a. a. watson , rev .
phys . * 72 * , 689 ( 2000 ) ; m. kachelrie , nucl .
* 165 * , 272 ( 2007 ) .
j. abraham _ et al .
_ [ pao collaboration ] , science * 318 * ( 2007 ) 938 ; astropart . phys .
* 29 * ( 2008 ) 188 [ erratum - ibid . *
30 * ( 2008 ) 45 ] .
israel , astron .
astrophys .
* 8 * , 237 ( 1998 ) .
see also http://www.mpe.mpg.de/cen-a/ . | we calculated for the nearest active galactic nucleus ( agn ) , centaurus a ( cen a ) , the flux of high energy cosmic rays and of accompanying secondary photons and neutrinos expected from hadronic interactions in the source .
we used as two basic models for the generation of ultrahigh energy cosmic rays ( uhecr ) shock acceleration in the radio jet and acceleration in the regular electromagnetic field close to the core of the agn , normalizing the uhecr flux to the observations of the auger experiment .
here we compare the previously obtained photon fluxes with the recent data reported by the fermi lat and h.e.s.s
. collaborations . in the case of the core model ,
we find good agreement both for the predicted spectral shape and the overall normalization between our earlier results and the h.e.s.s .
observations for a primary proton energy @xmath0 with @xmath1 or smaller .
a broken - power law with high - energy part @xmath2 leads to photon fluxes in excess of the fermi measurements .
the energy spectrum of the photon fluxes obtained by us for the jet scenario is in all cases at variance with the h.e.s.s
. and fermi observations . |
suspensions of dust , impurities , droplets , bubbles , and other finite - size particles advected by incompressible turbulent flows are commonly encountered in many natural phenomena and industrial processes .
these finite - size particles , whose density may differ from that of the underlying fluid , can not be modeled as point - like tracers because of their inertia .
often , they are characterized by the presence of strong inhomogeneities in their spatial distribution .
indeed , light ( heavy ) particles tend to concentrate into specific regions of the flow characterized by high ( low ) values of the vorticity .
such a phenomenon is dubbed ` preferential concentration ' @xcite .
the inhomogeneities then appearing in the particle spatial distribution are important because they affect the probability to find close particle pairs and thus influence their possibility to collide , or to have biological and chemical interactions .
examples showing the importance of this phenomenon are the formation of rain drops by coalescence in warm clouds @xcite , or the coexistence between several species of plankton in the hydrosphere @xcite .
engineering applications encompass optimization of spray combustion in diesel engines @xcite and in rocket propellers @xcite .
inertial particles are also important for the problem of dispersion of dust , chemicals or aerosols in the atmosphere @xcite .
recently , much effort have been devoted to the study of phenomena related to preferential concentration of inertial particles in turbulent flows by means of both theoretical @xcite and computational @xcite approaches .
progresses in the characterization of the statistical features of particle clusters have been achieved by studying inertial particles evolving in laminar stochastic flows @xcite and two dimensional turbulent flows @xcite .
experimental results are reviewed in ref .
recently , experiments @xcite have been also realized borrowing techniques from lagrangian turbulence @xcite .
considerable less attention has been paid to other aspects of inertial particle dynamics , and in particular to the statistical properties of their acceleration in turbulent flows .
this is an important issue relevant , for example , to the development of stochastic models of particle motion @xcite .
it is also interesting to contrast the statistics of the acceleration of inertial particles with that of fluid tracers which has been extensively studied by means of experiments @xcite , numerical simulations @xcite and theoretical approaches @xcite .
this is even more interesting on the light of recent experiments @xcite designed to study inertial particles in turbulent flows by means of lagrangian measurements .
concerning fluid tracers , it has been found that the acceleration probability density function ( pdf ) displays the typical fat tails of highly intermittent signals @xcite .
the tails of the pdf have been associated to trapping events into vortex filaments @xcite .
moreover , a model based on the multifractal description of turbulence has been proved able to describe the pdf of the acceleration tracers @xcite .
similar questions are still open for inertial particles .
because of the centrifugal force acting on such particles , inertia is expected to considerably change the statistics of trapping events into vortex filaments .
it is important to quantify this effect by looking at the modification of the pdf tails as a function of the degree of inertia .
moreover , it is not known how to generalize the multifractal approach used for passive tracers to the case of passive particles with inertia . in this paper
we focus on the statistical properties of particles much heavier than the carrier fluid .
we are interested in investigating , by means of high - resolution direct numerical simulations ( dns ) , the combined effects of preferential concentration and of inertia in determining the statistics of acceleration . with this aim ,
we review and extend the results presented in ref . @xcite .
in particular , we perform a systematic study as a function of the stokes number .
here stokes is defined as @xmath4 , being @xmath5 the typical response time of the particle and @xmath6 the kolmogorov time of the flow . + most results concern monodisperse suspensions , i.e. set of particles with a unique stokes number .
since laboratory and natural suspensions are typically polydisperse , we shall extend our analysis to this case also .
+ understanding the relevant time scales to reach a stationary regime is important both for experimental studies often limited in running time and for a dynamical description of preferential concentration .
we present an analysis of the _ transient regime _ ,
i.e. the time window necessary for the particles to reach a statistically steady regime , starting from their initial configuration .
the paper is organized as follows . in sec .
[ sec : hdyn ] , we recall the equations of motion for the case of inertial particles much heavier than the surrounding fluid . a detailed description of the numerical set up is given in sec .
[ sec : numerics ] . in sec .
[ sec : stat ] the validity of the model equations for our numerical experiments is considered .
moreover , we discuss the statistical features of preferential concentration both during the _ transient _ and in the statistically stationary regime .
the most relevant features of particle acceleration statistics for monodisperse suspensions are discussed in sec .
[ sec : accel ] .
section [ sec : poly ] is devoted to extend the previous analysis to polydisperse cases .
conclusions and perspectives are drawn in sec .
[ sec : con ] .
the equations of motion of a small , rigid , spherical particle immersed in an incompressible flow have been consistently derived from first principles in refs .
@xcite under a certain number of assumptions .
the main working hypothesis is to consider low particle concentration , so that it is possible to neglect collisions , particle - to - particle hydrodynamic interactions and the feedback of particles on the carrier fluid .
another hypothesis is that the particle radius @xmath7 is much smaller than any active scale of the turbulent flow , namely that @xmath8 , where @xmath9 denotes the kolmogorov dissipative scale .
finally , the particle reynolds number @xmath10 , where @xmath11 is the typical relative velocity of the particle with the flow and @xmath12 is the fluid kinematic viscosity , has to be small , i.e. @xmath13 .
these hypotheses lead to describe the fluid flow surrounding the particle as a stokes flow and to write a close differential equation for the motion of the particle @xcite .
the dynamics of a single particle then depends on only two dimension - less parameters .
the first is @xmath14 , accounting for the mass density ratio between the particle , @xmath15 , and the fluid , @xmath16 .
the second control parameter is the stokes number @xmath17 , where the particle response time is @xmath18 .
further , neglecting gravity , and assuming that the particle is much heavier than the fluid , the particle velocity @xmath19 is to leading order a solution to @xcite : @xmath20 where @xmath21 denotes the particle position , @xmath22 the fluid velocity field , and @xmath23 is the fluid acceleration at the particle position . as long as @xmath24 does not become too large ( see sect . [ subsec : validity ] ) , the first term in the right - hand side of ( [ eq:0 ] ) can be neglected when @xmath25 , so that the equations can be further simplified to @xmath26 it is interesting to compare the evolution equations for particles ( [ eq:1 ] ) with that for the motion of fluid tracers @xmath27 which corresponds to the limit @xmath28 of ( [ eq:1 ] ) .
tracers may be seen as @xmath2 particles .
the carrier fluid is evolved according to the incompressible navier - stokes equations @xmath29 where @xmath30 is the pressure field and @xmath31 is the external energy source , @xmath32 .
these are solved on a cubic grid with periodic boundary conditions .
energy is injected by keeping constant the spectral content of the two smallest wavenumber shells @xcite .
the viscosity is chosen so to have a well resolved dissipative range , @xmath33 where @xmath34 is the grid spacing .
we use a fully dealiased pseudospectral algorithm with 2@xmath35 order adam - bashforth time stepping .
we performed three sets of runs at resolution @xmath36 with @xmath37 , the corresponding reynolds numbers are @xmath38 ( see also ref .
@xcite ) .
particle dynamics , eqn .
( [ eq:1 ] ) , is integrated in parallel with that of the fluid by means of a 2@xmath35 order adam - bashforth time - stepping , in which the fluid velocity at particle position is estimated by linear interpolation . for each set of runs ,
the following procedure has been adopted .
we integrate the navier - stokes equations until the flow reaches a statistically steady state .
once a turbulent configuration for the flow is obtained , millions of particles and tracers are seeded in the flow .
their initial positions are uniformly distributed in the volume , and velocities are chosen equal to the local fluid velocity .
equations ( [ eq:1 ] ) for particles and ( [ eq : tracer ] ) for tracers are then advanced in parallel with those for the fluid ( [ eq:2 ] ) .
we consider particles with @xmath39 different stokes numbers ranging from @xmath40 to @xmath41 ( see caption of table [ table:1 ] for the exact values ) , and a set of tracers , @xmath2 .
for each set of particles with a given stokes number , we save the position and the velocity of @xmath42 particles every @xmath43 , with a maximum number of recorded trajectories of @xmath44 for the highest resolution . along these trajectories
we also store the velocity of the carrier fluid . at a lower frequency ,
@xmath45 , we save the position and velocity of a larger number @xmath46 of particles ( up to @xmath47 per @xmath48 , at the highest resolution ) together with the eulerian velocity field .
a summary of the physical parameters for the run at the highest resolution , the one discussed in this paper , is given in table [ table:1 ] .
[ cols="^,^,^,^,^,^,^,^,^",options="header " , ] [ tab : fit ]
direct numerical simulations , even at moderate reynolds number , represent a valuable tool for the study of the lagrangian motion of heavy particles .
here we have described the specific set up of our numerical study , analyzed the importance of preferential concentration and particle clustering as a function of the stokes number . moreover , we have explored the transient dynamics which precedes the settling of the particle motion onto a statistically steady state in terms of the particle distribution
. this may be particularly important for real observations and laboratory experiments , where long time records of particle motion are hardly attainable .
we have also reviewed and extended some recent results about the acceleration pdf of inertial particles , at varying the importance of inertia .
the main conclusions are ( i ) preferential concentration plays an _ almost singular _ role at small stokes , since even a small inertia is sufficient to expel particles from vortical regions where the strongest acceleration fluctuations are experienced ; ( ii ) for small stokes , a good quantitative agreement between the inertial particle acceleration and the conditioned fluid tracer acceleration is obtained ; ( iii ) at large stokes , the main effects is filtering of the velocity induced by the response stokes times . + the formulation of a phenomenological model able to describe the inertial particle acceleration as a function of both stokes and reynolds numbers is still beyond our reach . an easy fitting procedure has however been proposed for polydisperse suspensions , the most common experimental situation . for polydisperse cases ,
we show that the simple approach of characterizing the suspension in terms of its average stokes number without taking into account the stokes number distribution can give a systematic bias , particularly if the particle distribution is skewed towards small stokes values .
it would be extremely important to test the proposed approach with experimental data .
finally it would be interesting to numerically investigate @xmath49 particles to get a better understanding of the almost singular behavior that has been observed for the acceleration .
we acknowledge useful discussions with s. ayyalasomayajula , e. bodenschatz , a. gylfason , g. falkovich and z. warhaft . this work has been partially supported by the eu under the research training network hprn - ct-2002 - 00300 `` stirring and mixing '' .
numerical simulations have been performed thanks to the support of cineca ( italy ) and idris ( france ) under the hpc europa project ( rii3-ct-2003 - 506079 ) .
we thank also the `` centro ricerche e studi enrico fermi '' and n. tantalo for technical support . | we present the results of direct numerical simulations ( dns ) of turbulent flows seeded with millions of passive inertial particles .
the maximum reynolds number is @xmath0 .
we consider particles much heavier than the carrier flow in the limit when the stokes drag force dominates their dynamical evolution .
we discuss both the _ transient _ and the _ stationary _ regimes . in the transient regime , we study the growth of inhomogeneities in the particle spatial distribution driven by the preferential concentration out of intense vortex filaments . in the stationary regime , we study the acceleration fluctuations as a function of the stokes number in the range @xmath1 $ ] .
we also compare our results with those of pure fluid tracers ( @xmath2 ) and we find a critical behavior of inertia for small stokes values . starting from the pure monodisperse statistics we also characterize polydisperse suspensions with a given mean stokes , @xmath3 . |
the soft x - ray transient ( sxt ) 4u1543 - 47 was discovered on 17 august 1971 by the _ uhuru _ satellite @xcite . since then , additional x - ray outbursts have been observed in 1983 , 1992 and 2002 .
the optical counterpart was found during the 1983 outburst by @xcite .
dynamical studies of the companion star @xcite show that the compact object is most likely a black hole with a mass of 2.7 @xmath0 m@xmath1 7.5 m@xmath2 .
the 2002 outburst began on mjd 52442 ( ut june 16.683 ) and was observed by rxte @xcite . during this
outburst the soft x - ray light curve exhibited a classical fast rise , exponential decay profile as observed in many other soft x - ray transients @xcite .
the top panel in figure 1 shows the rxte asm light curve from mjd 52390 to 52500 . from early on in the outburst
4u1543 - 47 was in a thermal - dominated @xcite state ( also known as high - soft state ) .
between mjd 52457 - 52459 the object briefly transitioned into a low - hard state .
after this time , 4u 1543 - 47 went back into a thermal - dominated state until mjd 52476 when the source transitioned into the low - hard state @xcite and remained so until it reached quiescence .
we observed the 2002 outburst of 4u 1543 - 47 as part of the yalo ( yale , aura , lisbon , ohio ) consortium .
our observations cover most of the outburst cycle in ir and some in optical .
a `` secondary maximum '' appeared in the optical and ir ( oir ) light curves which is not apparent in the asm lightcurve . in this paper
we present our oir lightcurves and focus our analysis on the secondary maximum .
we argue that this secondary maximum is synchrotron emission which originates from a jet and that this may be a common phenomenon of sxts during outburst decline .
@xmath3- and @xmath4-band images of 4u1543 - 47 were obtained on a daily basis ( weather permitting ) from mjd 52323.3 ( ut 2002 february 17 ) using the yalo 1.0 m telescope with andicam , a dual - channel imager capable of obtaining oir data simultaneously .
data was recorded by a lick / loral-3 2048x2048 ccd on the optical channel and a rockwell 1024x1024 hgcdte `` hawaii '' array on the infrared channel .
on mjd 52431 ( ut 2002 june 5 ) the yalo optical ccd failed .
however , @xmath4-band observations continued on a daily basis . when the @xmath4-band light curve was observed to have risen by @xmath5 0.5 mag ( @xmath5 mjd 52440 ) we initiated additional observations in the @xmath6-band .
optical observations were obtained on a daily basis , when possible , from mjd 52442.9 - 52500.8 ( ut 2002 june 18 - august 15 ) using the 74 inch telescope at mount stromlo observatory with the cassegrain imager , a 2k x 4k ccd and @xmath7 and @xmath8 filters .
infrared observations ceased when the object was no longer high enough in the night sky to observe .
oir data reduction and photometry were performed using the usual routines in iraf .
calibration of the optical data was derived by calculating the offsets from a secondary standard , the magnitude of which was provided by jerry orosz ( private communication ) .
infrared images were calibrated via observations of the primary standard p9187 , observed on mjd 52520 ( ut 2002 september 2 ) .
the optical magnitudes were corrected for airmass extinction using coefficients taken from the ctio extinction table in iraf while the infrared atmospheric extinction coefficients were taken from @xcite .
figure 1 shows the oir light curves together with the rxte asm light curve .
the @xmath4-band data rises @xmath5 3 - 4 days before the asm , however , this difference will most likely decrease if the analysis is done with the more sensitive pca data instead .
it is not clear that there is any noticeable difference in the peak times between the asm and @xmath4-band .
after the primary x - ray maximum , the x - rays , oir lightcurves proceed on a steady decline . at mjd @xmath5 52460 the @xmath6-band lightcurve , and
to a lesser extend the @xmath4-band , experiences a small reflare .
it is interesting to note that just before this reflare 4u 1543 - 47 briefly transitioned into a low state and low - frequency qpos were observed @xcite .
the object went back into a thermal - dominated state at mjd 52463 . at mjd 52476 , 4u 1543 - 47 transitioned again into the low state when , again , low - frequency qpos were observed @xcite .
no optical or infrared data was obtained during the time of transition , however , soon after a third reflare is observed peaking at mjd 52487 .
this third reflare , which we call a secondary maximum , is much stronger in the infrared than in the optical and the strength of the peak increases with wavelength .
the reflare is so strong in the @xmath6-band that it s peak surpasses that of the primary maximum .
interestingly , no response is seen in the asm lightcurve .
this is not to say that there is no response in the x - rays at all .
indeed , preliminary analysis of the pca data shows a large decrease in the power - law index , qpo frequency and disk blackbody flux occurring close to the time the secondary maximum begins to rise @xcite .
a radio detection was made at the peak of the secondary maximum at a flux density of 3.8 @xmath9 0.6 mjy at 843 mhz ( jeff mcclintock , private communication ) .
after the secondary maximum peak , all lightcurves decline .
the @xmath4- and @xmath6-bands eventually follow the decrease from the primary peak until reaching quiescent levels .
the most extraordinary feature of these lightcurves is the secondary maximum at mjd 52487 and the strength of this feature in @xmath4 and @xmath6 compared to the optical bands .
this secondary maximum flux is in addition to the flux from the primary maximum decline since the infrared lightcurves return and follow the primary outburst decline after the reflare . to ascertain the origin of the secondary maximum we fitted a single blackbody , multicolor blackbody @xcite and broken power law to the spectral energy distribution of the secondary maximum .
the underlying flux from the primary outburst was removed by fitting a straight line to the data before and after the secondary maximum and subtracting the resulting fit .
hence , the sed shown in figure 2 is derived solely from the secondary maximum flux . in figure 2
we also show our best fits of each model to the sed .
the long - dashed line is a single blackbody model of t=1600k .
this model can account for the infrared data but none of the optical . at this temperature
, the emitting region would have a radius of @xmath5 2 x 10@xmath10 cm .
for an orbital period of 1.123 days @xcite , the primary roche lobe is @xmath5 5 x 10@xmath11 cm .
hence , if the secondary maximum emission is thermal , the source would be at a distance 10@xmath12 times further than the primary roche lobe , which seems implausible . = -0.78 @xmath9 0.14 .
short - dashed line shows power - law with @xmath13 = 0.062 @xmath9 0.003 , taken to be a lower limit . ]
the dot - dashed line in figure 2 represents our best multicolor - blackbody fit .
the fit parameters are as follows : @xmath14 = 10@xmath15 ( where @xmath16 was fixed to 3 gravitational radii ) , @xmath17k and m_dot = 10@xmath18 gs@xmath19 ( see mitsuda et al .
1984 for parameter definitions ) .
we fixed the mass of the black hole at 5 m@xmath2 @xcite .
as for the single blackbody case , the model can account for some but not all of the optical / infrared data and it is clear that the outer temperature given by the best fit is too cold for a physical disk . the broken power - law fit ( @xmath20 ) is shown as a dashed - solid line in figure 2 .
we use this broken power - law to represent emission from a synchrotron source .
the solid line is a linear least - squares fit ( in log space ) to the oir data , giving @xmath13 = -0.78 @xmath9 0.14 .
since we have only one radio datapoint , and since the sed must break somewhere between the radio and @xmath6-band points , we can only give a lower - limit of @xmath13 @xmath21 0.062 @xmath9 0.003 to the first part of the sed .
this is shown as the dashed line in figure 2 .
the broken power - law fit is much improved over the blackbody fits .
since the oir data are best described by the power - law fit , together with the unphysical parameters given by the blackbody fits , we conclude that the most likely source of the reflare is jet synchrotron emission .
this conclusion is supported by the radio detection made at the peak of the secondary maximum .
this is not the first time that this type of behavior has been observed during the outbust decline of soft x - ray transients .
@xcite observed a large reflare in the oir during the 2000 outburst decline of xte j1550 - 564 .
recent observations made by us @xcite of the 2002/2003 outburst of gx 339 - 4 ( see figure 3 ) show the same phenomenon . in both cases
, it appears after the sources transition into the low state and the reflare is much stronger in the infrared than optical .
past studies of sxt outbursts report a `` plateau '' in optical data during outburst decline ( see * ? ? ?
* for example )
. these plateau may be the same phenomenon as we see here .
it is now widely accepted that jets are a near - ubiquitous property of sxts during the low state @xcite .
this scenario is now increasingly supported by oir observations such as those presented in this paper .
indeed , since the infrared data show such strong reflares , they may be used as a reliable trigger for other multiwavelength observations . in doing
so we will be able to better constrain when jets appear and make progress in our understanding of jet formation and how this relates to the accretion geometry .
the authors would like to thank alfred chen for assisting in the optical data reduction and emrah kalemci and nick morgan for their careful perusal of the manuscript .
mb is supported by nsf grant ast-0098421 . | we have taken optical and infrared observations during the 2002 outburst of the soft x - ray transient , 4u 1543 - 47 .
a secondary maximum occurs in the lightcurves during the outburst decline .
this feature is much stronger at infrared wavelengths than optical .
we have applied single blackbody , multicolor blackbody and broken - power law models to the optical / infrared spectral energy distribution of the secondary maximum and find that the broken power - law provides the best fit .
we therefore conclude that the secondary maximum emission originates from a jet .
we also show the most recent lightcurves of the 2002/2003 outburst of gx 339 - 4 in which a secondary maximum appears .
this leads us to the conclusion that secondary maxima may be a common occurrence in soft x - ray transients during outburst decline which appear after the object transitions into the low - hard state .
infrared observations of such phenomena will give reliable triggers for multiwavelength observations , allowing us to greatly improve our knowledge of jet formation and behavior , and how this relates to the accretion geometry .
address = department of astronomy , yale university , new haven ct 06511 , u.s.a .
address = department of astronomy , yale university , new haven ct 06511 , u.s.a . |
strongly correlated phases of electrons confined to move in the plane , subject to a perpendicular magnetic field , have attracted significant attention since the discovery of fractionally quantized hall conductivity @xcite .
the profound role of topology in this extreme quantum limit leads to the presence of quasiparticles that carry a fraction of electron charge @xcite , and fractional ( abelian , or possibly non - abelian ) statistics @xcite .
the prospect of excitations possessing non - abelian statistics has motivated different schemes for topological quantum computation @xcite based on these systems .
these remarkable phenomena occur in the fractional quantum hall ( fqh ) regime , when the number of electrons , @xmath2 , is comparable to the number of magnetic flux quanta @xmath3 through the two - dimension electron system ( 2des ) .
correlated fqh liquid states appear at certain partial filling @xmath4 of the active landau level ( ll ) . in traditional semiconductor heterostructures ,
the physics of a partially - filled @xmath5 ll differs significantly from that of @xmath6 ll , due to the node in the single - particle wavefunction @xcite . as a consequence , the hierarchy / composite fermion states @xcite , ubiquitous in the lowest landau level ( lll ) ,
are significantly weakened in @xmath5 ll , and some of the more exotic states , such as the read - rezayi ( rr ) parafermion states @xcite , are likely to be favored . a number of studies have focused on the simplest , @xmath7 non - abelian member of the rr sequence the moore - read ( mr )
pfaffian " state @xcite .
the pfaffian , or perhaps its particle - hole conjugate version the anti - pfaffian @xcite , with the same topological properties , are widely believed to describe the fqh plateau at @xmath8 @xcite .
quasiparticles of the mr and higher rr states obey the non - abelian statistics @xcite which is of interest for topological quantum computation @xcite .
however , for the purpose of _ universal _ topological quantum computation , mr state is not sufficient , and one must go to a higher , @xmath0 member of the rr sequence ( see ref .
@xcite and for a general case ref .
@xcite ) . in gaas systems ,
experiments @xcite have detected only a weak plateau at @xmath1 , which theoretical work @xcite has tentatively identified with the particle - hole conjugate of the @xmath0 rr state .
this state , however , is very fragile and has been seen only in a small number of samples , triggering alternative theoretical proposals for its origin @xcite .
numerical calculations suggest that the appearance of the rr @xmath0 state is linked to the finite - width of the 2des in gaas @xcite , but little is known about the stability of @xmath0 rr state as the interaction is tuned away from the bare coulomb point .
for example , in case of @xmath7 ( moore - read ) state , it is possible to construct a 2-body interaction @xcite , resulting from particle - hole symmetrization of the hard - core 3-body repulsion , which yields a ground - state with high overlap with the exact ground state of coulomb interaction .
this model interaction consists only of @xmath9 and @xmath10 haldane pseudopotentials ( to be defined below ) . in case of @xmath0 rr state ,
such an approximate 2-body hard - core interaction has not been constructed , but one could expect that it requires higher order haldane pseudopotentials ( and perhaps some 3-body terms as well ) .
one of the main disdvantanges of the gaas - based devices is that their 2des is buried inside a larger , three - dimensional structure .
this unfortunate fact fixes the effective interactions at values that are often not optimal for some of the most interesting fqh states , including the rr series .
for example , the mr state is found to lie very close to the boundary with a compressible phase @xcite .
another problem stems from the strong dielectric screening and finite well - width @xcite in gaas , which weaken the electron - electron interactions and make fqh states fragile .
this has been a major obstacle in the studies of the non - abelian states , which could only be observed in ultra - high - mobility samples @xcite .
thus , it is desirable to find an alternative high - mobility 2des with strong effective coulomb interactions that are adjustable in a broad range .
recently , a new class of such high - mobility 2des which host chiral excitations with non - trivial berry phases , has been discovered .
these chiral " materials include graphene and bilayer graphene @xcite , and , more recently , topological insulators @xcite , as well as certain quantum wells @xcite .
the chiral nature of the quasiparticles gives rise to new electronic properties , including an unusual ll sequence , anomalous hall effect , and suppression of weak localization @xcite . when these chiral materials are subject to a perpendicular magnetic field , the kinetic energy quenches into discrete landau levels , similar to the usual semiconductors with non - chiral carriers .
however , due to the chiral band structure @xcite and the fact that the surface of these materials is exposed @xcite , they offer new possibilities to tune the effective interactions and explore strongly correlated phases . in this paper
we analyze two practical ways of tuning the effective interactions in chiral 2des , and the effect this has on the non - abelian fqh states . as a case study , we choose to focus on the @xmath0 rr state . in a stark contrast to gaas
, we find that the chiral 2dess allow for a more robust rr state along with the possibility of its realization in _ several _ lls .
additional insights can be obtained by driving transitions between the rr state and the abelian hierarchy state , or the compressible stripe phase @xcite .
such transitions can be implemented in chiral and _ massive
_ 2des by varying the external field .
overall , the tunability of the chiral 2des allows one to explore a larger region of the effective interactions than has been achieved in gaas .
we consider a family of 2d materials that are characterized by non - trivial berry phases .
one such material is monolayer graphene , a high - mobility atomically thick 2des @xcite , where recently several fqh states of the type @xmath11 have been discovered @xcite . a closely related material
, bilayer graphene @xcite has similarly high mobility , and exhibits interaction - induced quantum hall states at integer filling factors at low magnetic fields @xcite .
graphene and its bilayer are characterized by berry phase @xmath12 @xcite ( graphene - like ) and @xmath13 @xcite ( bilayer - graphene - like with an energy gap ) , respectively .
much of previous theoretical work on chiral materials has been restricted to graphene , exploring the consequences of the four - fold ll degeneracy ( valley and spin ) that leads to new su(2 ) and su(4)-symmetric fractional states @xcite .
instead , we focus on the high - field limit , neglecting the multicomponent degrees of freedom , and examine the effects originating from the interplay of the coulomb interaction and band structure .
we consider a family of band structures introduced in ref .
@xcite , which describe a number of high - mobility materials , including graphene with _ massive _ carriers ( mass is generated either spontaneously , or as a result of sublattice symmetry breaking @xcite ) , topological insulators @xcite , bilayer - graphene @xcite , trilayer graphene @xcite , and similar materials .
pristine graphene , which hosts massless dirac - like fermions , is contained in this model as a particular case . of the haldane pseudopotentials as a function of system s parameters : the width @xmath14 in case of the infinite quantum well model ( a ) , the band structure parameter @xmath15 ( b ) , and screening parameter @xmath16 and distance of the dielectric plate @xmath17 ( c ) .
chiral 2des offer the possibility to tune @xmath18 in a much broader range than it is possible within a quantum well model . ] more explicitly , we consider a family of @xmath19 hamiltonians with berry phase @xmath12 and @xmath13 @xcite .
the case of @xmath12-carriers is realized in graphene , topological insulators , special quantum wells @xcite ; the case of @xmath13-carriers occurs in bilayer graphene , where the energy gap can be controlled in a wide range by a perpendicular electric field @xcite . these @xmath19 effective hamiltonians can be derived pertubatively starting from a tight - binding model for each of the materials with a given chirality @xmath20 , however the effective model will be applicable in a narrower energy range for larger @xmath20 @xcite . as it is well known for the case of graphene ,
the one - body eigenstates of such hamiltonians possess a spinor structure .
consequently , the harmonic oscillator raising / lowering operators will mix the different components to produce an effective form - factor that has a richer structure than in case of a simple parabolic band such as in gaas , making it possible to tune the effective interaction . for @xmath12 carriers ,
the single - particle wavefunctions are given by @xcite @xmath21 , where @xmath22 is the wave function of the @xmath23th non - relativistic ll ( @xmath24 ) , and parameter @xmath15 depends on the ratio @xmath25 , where @xmath26 is the mass gap , @xmath27 the fermi velocity , and @xmath28 the magnetic length .
we use the notation @xmath29 to denote the @xmath23th ll for @xmath30 carriers ( @xmath31 ) . as a consequence of the spinor wavefunction , the effective form factor @xcite @xmath32 that describes the interaction projected to a @xmath33 ll
is given by @xmath34 where @xmath35 is the @xmath36th laguerre polynomial , and for simplicity we omitted the index of @xmath15 .
the form - factor is a mixture of the @xmath37th and @xmath38th ll form - factors in a non - relativistic 2des with parabolic dispersion . at @xmath39 ,
the above equation reduces to the form - factor of graphene @xcite , however , quite generally by varying @xmath25 , one can realize any value of @xmath40 .
similarly , for carriers with berry phase @xmath13 , the single - particle wavefunctions are @xmath41 , and the form - factor is a mixture of standard @xmath37th and @xmath42th form - factors , @xmath43 the tunable form of the effective interactions eqs.([eq : formfactors],[eq : formfactors2 ] ) provides a way to engineer transitions between strongly correlated phases _ in situ _ by changing the field .
as pointed out by haldane @xcite , an interacting many - body system of electrons , confined to a partially filled ll , is defined by a finite set of numbers the haldane pseudopotentials .
for rotationally invariant systems , these numbers @xmath44 represent the amplitudes of a state of two electrons with the relative angular momentum @xmath45 .
the physics of the fqhe is determined by small-@xmath45 pseudopotentials , i.e. @xmath46 . the role of higher @xmath44 s is to enforce charge neutrality , but they do not affect the incompressibility . therefore , the ratio of the two strongest pseudopotentials , @xmath18 , can be a rough indicator of the expected many - body phases . in fig .
[ fig : pp](b ) we show this ratio for several types of charge carriers and lls .
the variation of @xmath18 as a function of @xmath15 is to be contrasted with fig.[fig : pp](a ) where the same quantity is plotted for an infinite gaas quantum well of width @xmath14 . in gaas
the width of the well can be tuned by electrostatic gates ; however , it invariably reduces @xmath18 , whereas the chiral 2des allow for both increase _ and _ decrease of @xmath18 , and in a wider range .
up to this point , we have discussed the change in the effective interaction @xmath47 that resulted from a modification of the single - particle wavefunctions , and therefore @xmath48 .
the second method , proposed in ref .
@xcite , allows one to directly change @xmath49 .
chiral materials , such as graphene , are often exposed to the environment , which allows for dielectric material to be deposited on top of them .
we consider a setup where graphene sample is situated in a dielectric medium with permittivity @xmath50 , and a semi - infinite dielectric plate with permittivity @xmath51 is placed at a distance @xmath52 away from the graphene sheet . the effective interactions between electrons in graphene change due to the surface charges induced at the boundary between dielectrics @xcite : @xmath53 the ratio @xmath54 controls the effective interactions within a partially filled ll .
however , the overall energy scale is also modified and this has an impact on the magnitude of the excitation gap .
the gap should be multiplied by a factor @xmath55 if comparison is to be made with gaas 2des .
again , an important advantage of this setup is that the interactions can be tuned _ in situ _ by varying the magnetic field @xmath56 , which modifies the ratio @xmath54 .
the consequences for the many - body system are illustrated in fig .
[ fig : pp](c ) . in the dielectric case , @xmath18 is found to have a substantially larger variation as a function of @xmath16 than the quantum well model or the tuning of the band structure through @xmath15 . in particular , for intermediate @xmath57 and negative @xmath16 , @xmath18 can be readily increased by @xmath58 compared to its bare value ( @xmath59 ) .
therefore , based on the behavior of @xmath18 , we expect that chiral 2des offer many possibilities of tuning the interactions . by increasing this ratio ,
we expect to stabilize the incompressible phases ; on the other hand , by reducing it , one can drive a transition to compressible phases , such as stripes and bubbles . in the following
, we will explicitly verify these statements in exact diagonalization calculations , and determine the stability of the @xmath0 rr state . in the operational sense
, one can define several criteria for the stability " of a state ; the more of those criteria are met , the more
stable " the state is .
for example , one can compare the overlap ( scalar product ) between an exact many - body ground state and the rr wavefunction ; if the overlap , defined below , is consistenly close to unity for a number of system sizes , we consider it as one of the indicators that the state is in the rr phase .
however , there are known examples of different phases of matter which possess ground states of high overlap with each other in finite systems @xcite ; conversely , a relatively poor overlap can not be taken as a definitive proof that a phase is not realized .
additional indicators to identify the phase of matter are , for example , the ground state degeneracy on the torus , and the shift on the sphere ( selected to minimize the total ground - state energy ) .
finally , the robustness of a phase is also determined by the magnitude of the gap for creating neutral and charged excitations .
all of these indicators should be assessed together in determining the nature of a phase being studied .
we consider an interacting , @xmath60-electron problem with a landau - level spectrum given above , on a compact surface threaded by @xmath3 magnetic flux quanta using exact diagonalization . in the study of fqhe , two kind of surfaces
are available that preserve the translational invariance of an infinite 2deg : sphere @xcite and torus @xcite .
the two choices of boundary conditions illustrate the specific features of a fqh state under investigation : on a sphere , the fqh state couples to the curvature of the manifold , which is characterized by the topological number called shift @xcite .
the shift produces a small offset between @xmath3 and the magnetic monopole , whose strength is denoted by @xmath61 , placed in the center of a sphere . as a consequence , different fqh states live " in different hilbert spaces and in principle can be directly compared only after extrapolation to the thermodynamic limit . on the other hand ,
the flat surface of a torus leads to a unique definition of @xmath3 for given @xmath60 and @xmath62 , so that different candidate states describing the same filling @xmath62 are all realized in the same hilbert space .
the caveats of this geometry are the additional geometric parameters , the angle and aspect ratio of the torus ; the hamiltonian of a finite - size system depends on these parameters and their specific values favor one fqh phase over the others .
thus , an analysis is slightly more involved but the advantage is that a ground - state degeneracy can be used to identify the phase .
for example , in order to detect a stripe phase , we use a rectangular geometry with a specific aspect ratio that best accommodates the stripe in one direction .
in contrast , on the sphere fqh states always appear as non - degenerate , zero angular momentum ground states .
rr state and the exact ground state in @xmath63 and @xmath64 lls as a function of @xmath15 , for systems of different sizes @xmath60 .
high overlap is found close to the bare coulomb interaction in @xmath5 non - relativistic ll . at @xmath65 ,
the ground state undergoes a change in shift , and a hierarchy state sets in ( right ) . for @xmath66 in @xmath64 ll
, the system undergoes an incompressible - compressible transition to a stripe phase . ]
we will mostly present data for the spherical geometry because this allows for a number of simple diagnostic quantities to be evaluated in a straightforward manner .
the spherical representation of broken - symmetry states , however , necessarily contains defects , and periodic boundary conditions have to be used in this case .
a number of useful insights can be inferred from the study of the energy spectrum ; in addition , we will use overlap calculations to compare an exact , many - body ground state @xmath67 , with a numerical representation of a trial wavefunction , @xmath68 .
the overlap is defined as a scalar product between two normalized vectors , @xmath69 .
if @xmath70 is consistently close to unity for a number of system sizes considered , we consider the trial wavefunction to be a faithful description of a fqh phase . from the knowledge of a ground - state wavefunction @xmath71 , we also evaluate the ( projected ) static structure factor @xcite .
sharp peaks in the structure factor indicate the onset of compressible phases @xcite .
rr state and the exact ground state in @xmath72 ll as a function of @xmath15 , for various system sizes @xmath60 .
the region of high overlap is found in the window between the stripe phase and the hierarchy phase . ] in fig .
[ fig : z3 ] we show the overlap between the @xmath0 rr wavefunction and the exact ground state of a system in @xmath63 and @xmath64 lls , as a function of @xmath15 .
we set the shift is @xmath73 , corresponding to that of the rr state , and do not consider other candidate states with different values of the shift @xcite .
rr model wavefunction can be obtained by diagonalizing a four - body short - range interaction @xcite , which quickly becomes numerically prohibitive as the number of particles grows , or recursively via jack polynomials @xcite .
the filling factor is fixed at @xmath74 , but in the absence of ll mixing , our results directly apply to @xmath75 via particle - hole symmetry .
the highest overlap is found for the effective interaction that is in the vicinity of the pure @xmath5 non - relativistic coulomb potential .
smaller systems ( under @xmath76 particles ) indicate that rr state might be enhanced for @xmath77 slightly less than 1 in @xmath64 ll , which corresponds to a small relative _ decrease _ of @xmath9 pseudopotential @xcite .
however , the largest system one can obtain by exact diagonalization , @xmath76 , suggests that the maximum overlap is practically at the bare coulomb point .
it can not be determined with certainty whether this is an intrinsic feature of the rr state or a finite size effect that plagues @xmath76 system .
finite - size effects are generally very strong at this filling factor and further theoretical studies , possibly employing density - matrix renormalization group techniques , are needed . on either side of the bare coulomb point , the rr state
is quickly destroyed .
on the left , for @xmath66 in @xmath64 ll , there is an incompressible - compressible transition to a stripe phase . as mentioned earlier , states with broken translational symmetry are not adequately represented in the spherical geometry ; however we can still detect the stripe by the presence of a sharp peak in the structure factor .
one arrives at the same conclusion ( but more transparently ) in the torus geometry , where it is found that the system develops a manifold of nearly degenerate ground states for @xmath66 that are characterized by a linear ( uni - directional ) array of pseudomomenta that defines the preferred direction of a stripe .
as we move away from the pure coulomb point in @xmath63 ll , the ground state undergoes a change in shift for @xmath78 in order to minimize the energy .
the new value of the shift is that of a hierarchy state .
therefore , in chiral 2des , one can in principle study a subtle phase transition between two topological phases with abelian and non - abelian excitations .
an interesting feature , shown in fig .
[ fig : z3_bilayer ] , is the presence of @xmath0 rr correlations in a @xmath72 ll . note that the effective interaction in this case interpolates between pure @xmath6 and @xmath79 ( non - relativistic ) ll coulomb repulsion ; therefore it is quite different from the @xmath5 ll form - factor ( fig.[fig : z3 ] ) . in this case , rr state appears in a narrow window , between the stripe phase ( for @xmath80 ) and the hierarchy state ( shaded region in fig .
[ fig : z3_bilayer ] , for @xmath81 ) .
we note that @xmath82 system has a significantly poorer overlap than all the other systems considered , including a larger @xmath76 system , which leads us to conclude that the poor overlap might be a finite - size effect .
more importantly , the excitation gap in the window where the ground state is well - described by the rr wavefunction is very small ( and also smaller than in @xmath63,@xmath64 lls ) .
this , together with the presence of a strong incompressible hierarchy phase , would make it difficult to observe experimentally . ) at @xmath74 for the generalized material described by eq .
[ eq : formfactor_general ] .
the gap smoothly decreases from the hierarchy phase down to zero ( stripe phase ) .
rr state appears in the transition region .
the gaps are calculated for @xmath82 particles at the shift @xmath73 . ] in the remainder of this paper , we investigate the dependence of the excitation gap as we tune the parameters of the system . as mentioned above , the experimental gap at @xmath1 is tiny .
it is therefore essential to find ways to enhance this gap in order to manipulate the state , e.g. via surface probes . to this end
, we consider a generalized model for the material that involves a superposition of @xmath6 , @xmath5 and @xmath79 ll non - relativistic form factors @xcite : @xmath83 in fig .
[ fig : chargegap ] we plot the charge gap at @xmath74 for @xmath82 particles .
the ground state of this particular system has a somewhat poorer overlap with the rr state than other system sizes , however the charge gap shows no such difference from other cases . along certain lines , indicated by arrows
, the phase diagram reduces to one of the cases studied above .
note that proper finite - size scaling needs to be performed in order to get the correct values for the gap ; however , it was previously found for filling factors @xmath84 and @xmath85 @xcite that this rigorous analysis produces values that are roughly in agreement with the ones shown in fig .
[ fig : chargegap ] . at @xmath74
the number of available system sizes is too small to perform such an extrapolation reliably , and the plot in fig .
[ fig : chargegap ] should only be viewed qualitatively .
we find a smooth transition between the hierarchy state and a stripe , with the gaps gradually dropping to zero for small @xmath77 . in the curved , narrow region around the transition
, we previously found large overlaps with the rr state @xcite .
the excitation gaps in the transition region are much smaller than in the hierarchy phase , but they are nonetheless higher than at pure @xmath5 non - relativistic ll point ( @xmath86 , @xmath87 )
. therefore , tuning the band structure does allow for a modest enhancement of the gap in the rr phase .
qualitatively similar results are obtained for the neutral , instead of charge , gap .
a larger variation of the gap can be achieved in the setup that utilizes the dielectric screening ( eq . [ eq : interaction_screened ] ) . in fig .
[ fig : dielectric ] we show the charge gap as a function of the screening parameter @xmath16 and the distance of the dielectric plate , @xmath54 . unless @xmath16 is close to unity
, the gap does not vary significantly with @xmath54 , which hints at the strong finite - size effects present even in this , fairly large , system of @xmath82 particles .
however , as @xmath16 becomes close to unity , the gap displays a clear maximum for @xmath57 , which is roughly two times larger than for the bare interaction .
the overlap with the rr wavefunction is zero for @xmath88 , but fairly large otherwise .
therefore , at least for positive @xmath16 , the gap can be tuned without reducing the good overlap with the model wavefunction . ) at @xmath74 as a function of the screening parameter @xmath16 and the distance of the dielectric , @xmath54 .
the gap is largest for @xmath16 close to unity and intermediate @xmath57 . for negative @xmath16
, the gap vanishes and the system is in the stripe phase .
data is shown for @xmath82 particles on the sphere at the shift @xmath73 . ]
in this paper we have explored the stability of @xmath0 rr state , the prime candidate for topological quantum computation , in the phase diagram as we tune the effective interaction via band structure parameters ( that depend on the chirality of the material ) , or via dielectric screening .
numerical calculations suggest that the rr state is most stable in the vicinity of the bare @xmath5 non - relativistic ll coulomb potential , the same one which describes gaas materials .
nevertheless , the chiral 2dess offer several advantages over gaas .
firstly , we find that rr state can be realized in various ( as opposed to a single ) lls , where the interactions have different form from the one in gaas .
more importantly , the excitation gaps can be enhanced and the nature of the rr state can be further probed by driving _ in situ _ quantum phase transitions , either to compressible states with stripe ordering , or to an incompressible ( hierarchy / composite fermion ) state .
this work was supported by doe grant de - sc@xmath89 .
y.b . was supported by the state of florida .
we acknowledge previous collaboration with r. thomale .
99 a. yu .
kitaev , ann . of phys . * 303 * , 2 ( 2003 ) ; _ ibid _ * 321 * , 2 ( 2006 ) ; m. freedman _
et al . _ ,
phys . * 227 * , 605 ( 2002 ) ; _ ibid _ * 228 * , 177 ( 2002 ) ; c. nayak _ et al .
phys . * 80 * , 1083 ( 2008 ) .
v. m. apalkov and t. chakraborty , phys .
lett . * 97 * , 126801 ( 2006 ) ; c. tke and j. jain , phys .
b * 75 * , 245440 ( 2007 ) ; z. papi , m. o. goerbig , and n. regnault , phys . rev
. lett . * 105 * , 176802 ( 2010 ) .
et al . _ , nat
. phys . * 7 * , 948 ( 2011 ) ; t. taychatanapat _
et al . _ ,
* 7 * , 621 ( 2011 ) ; a. kumar _
et al . _ ,
lett . * 107 * , 126806 ( 2011 ) ; c. h. lui _
et al . _ ,
* 7 * , 944 ( 2011 ) . | the @xmath0 read - rezayi ( rr ) parafermion quantum hall state hosts non - abelian excitations which provide a platform for the universal topological quantum computation .
although the rr state may be realized at the filling factor @xmath1 in gaas - based two - dimensional electron systems , the corresponding quantum hall state is weak and at present nearly impossible to study experimentally . here
we argue that the rr state can alternatively be realized in a class of chiral materials with massless and massive dirac - like band structure .
this family of materials encompasses monolayer and bilayer graphene , as well as topological insulators .
we show that , compared to gaas , these systems provide several important advantages in realizing and studying the rr state .
most importantly , the effective interactions can be tuned _ in situ _ by varying the external magnetic field , and by designing the dielectric environment of the sample .
this tunability enables the realization of rr state with controllable energy gaps in different landau levels .
it also allows one to probe the quantum phase transitions to other compressible and incompressible phases . |
the central regions of clusters of galaxies are very interesting and enigmatic places .
the intracluster medium ( icm ) in the central region is often so dense that its radiative cooling time is significantly shorter than the hubble time .
therefore , the icm will cool down unless it is heated significantly .
unbalanced radiative cooling would cause a `` cooling flow '' in the cluster center ( see fabian 1994 for a review ) .
x - ray imaging analyses by _ rosat _ and _ exosat _ indicated that mass deposition rates , @xmath1 , were more than @xmath2 for many clusters ( edge , stewart , & fabian 1992 ; allen & fabian 1997 ; peres et al .
however , the fate of the resulting cold gas , or whether icm in `` cooling flow '' clusters really cools down to low temperatures , is still unclear .
the cd galaxies in cooling flows have an excess blue star component , which suggests that the cooled icm forms new stars ( mcnamara & oconnell 1989 ) .
however , the implied star formation rates are much lower than the cooling flow mass deposition ones ( mcnamara & oconnell 1989 ) .
furthermore , recent _ xmm - newton _ high resolution spectroscopic observations have imposed a strong constraint on the amount of gas cooling down to low temperatures ( peterson et al .
2001 ; kaastra et al . 2001 ; tamura et al .
2001 ) , which is consistent with the fact that the mass deposition rates determined spectroscopically by _ asca _ tended to be lower than those determined by _ exosat _ and _ rosat _ ( see makishima et al .
2001 for a review ) .
the cd galaxies in the cooling flow clusters usually host strong radio sources , with a central radio core and/or radio lobes .
observations utilizing _ chandra _ s excellent spatial resolution have revealed interactions between these radio lobes and icm , which may affect the dynamical and thermal history of icm in the cluster center . in hydra a ( mcnamara et al .
2000 ; david et al .
2001 ) , perseus ( fabian et al .
2000 ) , and abell 2052 ( blanton et al . 2001 ) , radio lobes coincide with x - ray low surface brightness regions , which suggests that the radio plasma expands to displace the icm . on the other hand , x - ray cavities without radio emission are also found at larger distances in abell 2597 ( mcnamara et al .
they are believed to be old radio lobes which can not emit ( at least at high radio frequencies ) because the high energy relativistic electrons have already suffered radiative losses .
these radio quiet cavities also indicate that cd galaxies provided the icm with high energy electrons and magnetic fields intermittently .
abell 3112 is a cooling flow cluster at a redshift of @xmath3 .
there is a powerful radio source , pks 0316 - 444 , in the cluster center .
previous x - ray imaging analyses with _ exosat _
( edge , stewart , & fabian 1992 ) and _ rosat _ ( allen & fabian 1997 ; peres et al .
1998 ) indicated that abell 3112 had a strong cooling flow , with a mass deposition rate of @xmath4 @xmath5 yr@xmath6 .
the temperature profile at a larger scale than the cooling radius was studied by markevitch et al .
( 1998 ) and by irwin , bregman , & evrard ( 1999 ) with _ asca _ data .
markevitch et al .
( 1998 ) found that the temperature decreased outward from 6 kev to 3 kev with _
asca _ data , although they explicitly assumed a cooling flow component in the central region . on the other hand , irwin et al .
( 1999 ) argued that the data were consistent with an isothermal distribution .
an increase in the iron abundance toward the center was reported by finoguenov , david , & ponman ( 2000 ) . however , there was no detailed study of the temperature and abundance structure within the cooling radius ( @xmath7 kpc ) because of the limited spatial resolution of _
asca_. furthermore , any interaction region between the radio source and the icm would be too small to have been resolved with _
rosat _ or _ asca_. in this paper , we present _ chandra _ observations of central region of abell 3112 .
we assume @xmath8 km s@xmath6 mpc@xmath6 , @xmath9 , and @xmath10 . at a redshift @xmath3
, 1 corresponds to 1.94 kpc .
the errors correspond to the 90% confidence level throughout the paper .
abell 3112 was observed twice with _ chandra _ on 2001 may 24 for 7,257 seconds and on 2001 september 15 for 17,496 seconds .
each observation was pointed so that the cluster center would fall near the aimpoint of the acis - s3 detector . however , the roll angles were different , so the outer parts of the field of view differ for the two observations .
the two observations were merged , using the positions of bright point sources to register the two images . here
, we analyze data from the s3 chip only .
the data from the s1 chip was used to search for the background flares . a few short periods with background flares were found and removed using the lc_clean software provided by maxim markevitch , leaving a total exposure of 21,723 seconds .
only events with asca grades of 0 , 2 , 3 , 4 , and 6 were included .
standard bad pixels and columns were removed .
exposure maps and background data were generated for each observation separately , and then merged appropriately .
blank sky background data were taken from the compilation by maxim markevitch .
the data are possibly affected by the low energy qe degradation of acis .
we tried to correct this using the corrarfprogram .
however , the correction did not appear to work well for this data ; specifically , corrarf seemed to over - correct the data .
as a result , we obtained absorption column densities which were much lower than the galactic value in our spectral fits .
we expect that the absorbing column should be at least as large as the galactic value .
therefore , we will show the results without the correction for the low energy qe degradation .
fortunately , this correction significantly affects only the very low energy ( @xmath11 1 kev ) part of the _ chandra _ band . as a result ,
neither temperature nor abundance determinations were seriously affected by this correction . on the other hand , mass deposition rates of cooling flow components increased by a factor of about two when we fitted the spectra after the correction .
the observations were made with very large array at a center frequency of 1320 mhz on 1996 october 18 .
the array was in the ` a ' configuration , which provided an angular resolution of 6.9 @xmath12 1.5 arcsec .
a total of 1.2 hours were obtained on source using 32 channels across a 6.25 mhz band .
both right and left circular polarizations were observed .
phase and bandpass calibration was obtained by short ( 1 min ) observations of the strong ( 4.9 jy ) calibrator j0440 - 4333 taken every @xmath1320 minutes .
atomic hydrogen absorption was searched for towards the compact core over a velocity range from 22,10023,440 km s@xmath6 at a resolution of 44 km s@xmath6 , but not detected down to a 3@xmath14 limit on the maximum optical depth of 0.012 .
the x - ray image was adaptively smoothed using the ciao csmooth program with a minimum signal - to - noise of 3 per smoothing beam .
an identically - smoothed background was subtracted , and the image was divided by an identically - smoothed exposure map .
the exposure map was corrected for vignetting , using the typical energy of the observed photons to take account of the energy dependence of the vignetting .
the resulting image is shown in figure [ fig : corimage ] .
as in other typical cooling flow clusters , x - ray emission is distributed symmetrically and is fairly concentrated in the center .
the image is smooth and quite symmetric on large scales .
the cluster emission is elliptical , with a major axis along the direction from nne to ssw . on large scales
, there appears to be no substructure except for some point sources .
we used a wavelet detection algorithm ( wavedetect in ciao ) to detect individual point sources .
sources were visually confirmed on the x - ray image , and a few low - level detections were removed . using this method
, we found sixteen point sources .
one of these corresponds to the radio source pks 0316 - 444 , which is located at the center of the central cd galaxy eso 248- g 006 .
another corresponds to a galaxy in the cluster ( lcrs b031619.9 - 442441 ) .
the regularity of the cluster x - ray emission on large scales does nt continue into the very central region .
figure [ fig : center1 ] shows the x - ray contour map of the @xmath15 region around the central point source . in the very center , an elongated structure from the central point source towards the se is clearly seen .
we can see another filament - like component from @xmath16 south of the central point source towards the sw . to clarify structures which deviate from symmetric cluster emission , we fitted the entire cluster image with a concentric elliptical isophotal model , and then subtracted the best - fit model from the image to get the residual component . in the fitting procedure , we fixed the center of the isophotes to the center determined from the larger scale isophotes about which the image was symmetric , rather than the central point source .
the ellipticity and position angle for each isophote were allowed to vary .
it is most likely that relaxed cluster potential structure is not spheroidal but triaxial ( jing & suto 2001 ) . in this case
, the icm hydrostatic density distribution also will be triaxial ( lee & suto 2003 ) .
therefore , the ellipticities and the position angles of x - ray isophotes can vary with radius unless our line - of - sight coincides with one of the three symmetry axes . in figure
[ fig : center2 ] , the greyscale shows the residual produced by subtracting our best - fit elliptical isophotal model from the adaptively smoothed image of the central region of the cluster .
the sharp elliptical edge near the outside is an artifact because the elliptical model extended only to this isophote .
dark areas are positive residuals ( excess emission ) , while light areas are negative residuals .
the dark and light regions at the center are due to the central point source , which is not exactly at the center of the x - ray isophotes at large radii .
there are regions of excess emission to the south of the central source .
the contours show the 1.32 ghz vla radio image . there is a very bright radio core , which is unresolved in the radio image ;
the elliptical contours for the core show the beam of the radio observation .
the very bright radio core is coincident with the central x - ray source .
there are two diffuse radio regions to the se and sw of the radio core , which may be connected to the core .
the ns extent of these regions is presumably exaggerated due to the elongated observing beam .
roughly speaking , the excess x - ray regions appear to surround two lobe - like radio components .
however , part of the sw radio lobe is coincident with a region of x - ray excess emission .
unfortunately , the low resolution of the radio data , bright radio core , and elongated radio beam make it difficult to decide definitely whether the excess is in the radio lobes or partly surrounding it .
in order to examine the temperature and abundance structure quantitatively , we determined the x - ray spectra in annuli .
we fitted the data between 0.5 and 10.0 kev with a photo - absorbed single temperature mekal model ( kaastra 1992 ; liedahl , osterheld , & goldstein 1995 ) after masking the point sources .
we tried to fit the data fixing the absorbing column density to the galactic value ( @xmath17 @xmath18 ; dickey & lockman 1990 ) or allowing it to vary freely .
the fitting results depend somewhat on whether we fixed the absorption to the galactic value or not .
figure [ fig : radtemabund ] shows the radial temperature and abundance profiles .
the solid crosses are the values obtained when we allowed the absorption to vary , while the dashed crosses are the values when we fixed the absorption to the galactic value .
the dashed crosses are shifted slightly to the left in order to be more easily seen .
the results with fixed or varying absorption are similar . at @xmath19 , both the temperature and abundance are nearly constant at @xmath20 kev and @xmath21 , respectively .
the temperature decreases from @xmath22 kev at 70 to 3.5 kev in the central 20 .
the abundance increases from @xmath23 at 70 to 1.3 @xmath24 in the central 20 .
the temperature values obtained with variable absorption , along with their associated errors , are fitted to a power - law profile given by @xmath25 . for @xmath26 , @xmath27 kev and @xmath28 . for @xmath29 , @xmath30 kev and @xmath31 .
we will use these fitted temperature profiles later when we analyze conduction timescale and integrated gravitational mass . to examine the non - radial temperature structure
, we made a two - dimensional temperature map of the central @xmath32 region .
we divided the region into 64 ( @xmath33 ) square regions , each of which is @xmath34 .
then , we fitted the data of each region with the photo - absorbed mekal model where absorption is treated as a free parameter .
figure [ fig : tmap ] shows a two - dimensional temperature map of the central @xmath32 region overlaid with x - ray surface brightness contours .
black and white represent lower ( @xmath35 kev ) and higher ( @xmath22 kev ) temperatures , respectively .
there is no significant azimuthal structure in the temperature map while temperature decrease toward the center is again clearly seen .
we also made a two - dimensional abundance map in the same way .
however , abundances in the individual fits were too poorly constrained to provide a useful map .
we tried to fit the annular spectra with a model for a multiphase cooling flow to constrain the contribution from a cooling flow component and a mass deposition rate in each annulus , although the data within each annulus are almost consistent with a single temperature mekal model . as a cooling flow spectral model , we used the mkcflow model based on the mekal model .
we added a mekal model representing the emission from the icm outside the cooling flow .
both mkcflow and mekal were assumed to be subject to the same absorption column , which we fixed to the galactic value or let vary freely .
we fixed the metallicity and initial gas temperature of the mkcflow model to the same values of metallicity and temperature as the mekal model , respectively .
the low temperature in the mkcflow model is fixed to the lowest value ( 0.001 kev ) in xspec .
the results are shown in table [ tab : cflow1 ] and [ tab : cflow2 ] .
compared with the fitting results without cooling flows , the results are not improved dramatically .
figure [ fig : radmdot ] shows a radial profile of the mass deposition rate derived from the fitting .
the circles and solid error bars are the values obtained from fits with freely varying absorption .
the squares and dashed error bars are the values from the fits with fixed galactic absorption . when absorption is set as a free parameter , forth bin and the outermost three bins are consistent with a model without a cooling flow component .
only the inner three bins and the fifth bin have the cooling flow component , although their errors are fairly large and comparable to themselves .
the resultant total mass deposition rate is @xmath36 yr@xmath6 , which is significantly smaller than the rates derived from the previous image analyses of _ exosat _ and _ rosat _ data , which are @xmath37 yr@xmath6 ( edge , stewart , & fabian 1992 ; allen & fabian 1997 ; peres , et al .
the cooling flow component contributes less than @xmath38% of the total x - ray emission .
when we adopt the galactic absorption value , the mass deposition rate becomes even lower and consistent with a model without a cooling flow component in all bins .
even though multiphase gas is not found within small regions of the cluster , we do see an overall temperature gradient , with gas in the center that is cooler by a factor of @xmath39 compared with gas in the outer regions of the cluster .
therefore , if we fit a spectrum extracted from the whole cluster , a significant mass deposition rate and a low temperature cutoff might be expected .
we fit the data within @xmath40 with the mekal plus mkcflow model where the low temperature in mkcflow is allowed to vary .
the results are shown in table [ tab : totspec ] .
the upper and lower rows show the results with the absorption column allowed to vary and fixed to the galactic value , respectively .
in contrast to the spectral analyses in annuli , the mass deposition rate is comparable with the former values based on imaging analysis . however , the low temperature in the mkcflow model is not very low ( @xmath41 kev ) , which means that the observations are consistent with significant cooling , but only over a limited temperature range .
this is consistent with what has been observed with many cooling flow clusters using chandra and xmm - newton ( e.g. , peterson et al .
please note that the goodness of the fit is marginal ( the reduced chi - squared is about 1.5 ) .
this might indicate that the emission measure distribution with the temperature is not that of the standard cooling flow model even in the temperature range between @xmath42 and @xmath43 .
to quantify the radial structure of the icm , we made a radial profile of the x - ray surface brightness in the 0.3 - 10.0 kev band ( figure [ fig : radsfpden]a ) .
the bins for x - ray surface brightness were chosen to be the same as those used below to determine the radial variation in the x - ray spectrum ( [ sec : spectrum ] ) .
the surface brightness values were deprojected to determine the x - ray emissivity and gas density ( figure [ fig : radsfpden]b ) , assuming the emissivity is constant in spherical shells .
when we convert the emissivity into the gas density , we use a mekal code and assume that the temperature is constant in spherical shells and equal to that derived from the spectral fitting of projected data .
some fluctuations are seen in the density and pressure profiles near the outer boundary ( figure [ fig : radsfpden]b and [ fig : radsfpden]c ) , which are artifacts of the deprojection since we assume zero emissivity outside the outer boundary .
however , many simulations of our deprojection method show that this affects only a few outermost points , and has no significant effect anywhere near the center .
the observed gas density and pressure profiles are quite smooth outside of the region of the radio source . based on the smooth radial profiles and lack of substructure in the x - ray image outside of the radio source region ,
we conclude that this cluster is dynamically relaxed and that the icm is generally in hydrostatic equilibrium .
the gravitational mass distribution was determined from the equation of hydrostatic equilibrium , @xmath44 we use the fitted power - law temperature profiles when @xmath45 is calculated . to reduce the noise in the density gradient term , we determine the gradient at @xmath46 by differencing the densities at @xmath47 and @xmath48 .
uncertainties in the density gradient at @xmath46 are also calculated from the density uncertainties at @xmath47 and @xmath48 .
the density gradients and their uncertainties are assumed to be equal in the innermost three points and the outermost three points .
the gas mass profile is also determined from the icm density profile assuming spherical symmetry .
the gas mass is then given by @xmath49 the integrated gravitational mass and gas mass are shown in figure [ fig : radmrmg ] ( a ) . a radial profile of the gas mass fraction is shown in figure [ fig : radmrmg ] ( b ) .
the gas mass fraction increases from @xmath50 at @xmath51 to @xmath52 at @xmath53 .
this trend is similar to those in other well - relaxed clusters ( david , jones , & forman 1995 ; ettori , & fabian 1999 ; allen , schmidt , & fabian 2002 ) . our gas mass fraction is lower than that of mohr , mathiesen , & evrard ( 1999 ) at much larger radii ( @xmath54 1 mpc ) , but the gas fractions of clusters generally increase with radius , and this appears to apply to abell 3112 .
using the same beta - model fits as in mohr et al . , we find that the gas fraction at @xmath53 is predicted to be @xmath1312% , which agrees well with our values .
because the temperature is higher than @xmath35 kev over the whole cluster , the main emission mechanism is thermal bremsstrahlung .
therefore , the isobaric cooling time is ( sarazin 1986 ) @xmath55 radiative cooling would make a radial temperature gradient in the central region of the icm through its density dependence , as is observed .
[ fig : radtemabund ] ) . on the other hand
, thermal conduction would have the effect of reducing the temperature gradient .
the conduction timescale is generally expressed as ( see also sarazin 1986 ) @xmath56 where @xmath57 is electron number density , @xmath58 is the scale length of the temperature gradient , and @xmath59 is the boltzmann constant . if we consider only the coulomb scattering process , the thermal conductivity for hydrogen plasma is ( spitzer 1962 ) @xmath60 where @xmath61 , the coulomb logarithm , is @xmath62.\ ] ] figure [ fig : radtctcond ] shows radial profiles of @xmath63 and @xmath64 .
when we calculated @xmath64 , we used the fitted power - law temperature profile to reduce the noise in the temperature gradient .
also , @xmath64 is evaluated only for @xmath65 because an isothermal distribution is consistent with the temperature data in the outer regions .
the cooling time of the innermost bin is @xmath66 yr .
if we define a cooling radius as where @xmath63 is @xmath67 yr , it becomes @xmath68 , or @xmath69 kpc .
both results are consistent with the former _ rosat _ ( allen & fabian 1997 ; peres et al . 1998 ) and _ exosat _ results ( edge , stewart , & fabian 1992 ) . at the innermost three bins , @xmath63 and @xmath64
are comparable with each other . outside of this region ,
the conduction timescale is clearly shorter than the cooling timescale .
if conduction proceeded at the spitzer rate , it would erase the temperature gradient . since the observations show a significant gradient in this region ( fig .
[ fig : radtemabund]a ) , thermal conduction must be significantly suppressed by , e.g. , tangled magnetic fields and/or plasma instabilities .
the cooling rate of gas with solar abundance is only @xmath13 10% higher than that of hydrogen - helium mixture gas at 3.5 kev ( see figure 9 - 9 of binney & tremaine 1987 ) .
therefore , actual cooling timescales in the central regions are probably @xmath13 10% shorter at the most , which does not change the situation here dramatically .
although our estimation here is rough , it is still useful for an estimation of an order of magnitude .
calculations taking account of time evolution will be helpful to make more precise models .
we detected an x - ray point source in the cluster center .
the position is coincident with the optical core of the cd galaxy and the radio core .
the position of the x - ray point source is ( epoch j2000 ) @xmath70 @xmath71 with an uncertainty of 0.04 arcsec in each coordinate .
the spectrum of the point source is shown in figure [ fig : cpssp ] .
we fitted the data with a mekal model plus a power - law model .
in addition to the absorption column density common for both components , we added intrinsic absorption for the power - law component only . as a result ,
the applied model is @xmath72.\ ] ] the abundance of the mekal component was fixed to the value derived from the region surrounding the point source .
the absorbing column @xmath73 was also fixed either to the value derived from the surrounding region or the galactic value .
the fitting results are shown in table [ tab : centps ] .
the spectral model consists of a 1.26 kev thermal plasma component and a power - law component with a photon index of @xmath74 .
the power - law component has an extra absorbing column of @xmath75 @xmath18 .
this is consistent with the central source being a strongly absorbed agn .
we found that the central radio source associated with the central cd galaxy in abell 3112 is interacting with the surrounding icm ( fig .
[ fig : center1 ] and [ fig : center2 ] ) .
the radio image shows a central core and two diffuse lobes to the se and sw of the core .
the central x - ray image has an asymmetric structure , and the excess x - ray emission over an elliptical isophotal model roughly appears to surround two radio lobes .
this configuration is naturally explained if the radio lobes have swept up the icm which was where the radio lobes are now .
similar scenarios have been proposed in other cooling flow clusters with radio sources ( e.g. , mcnamara et al .
2000 ; fabian et al .
2000 ; blanton et al .
in addition , there is excess x - ray emission coincident with part of the sw radio lobe .
it is possible that the excess x - ray emission is due to the cool gas that was originally near the very center ; this cool gas might have been entrained by hot buoyant bubbles .
this scenario was originally proposed to explain the x - ray morphology of the central region of m87 ( churazov et al .
indeed , the filamentary structures in the x - ray image of the central region of abell 3112 ( figure [ fig : center1 ] ) are similar to those in m87 ( feigelson et al . 1987 ; bhringer et al . 1995 ) .
this model may also apply to abell 133 ( fujita et al .
2002 ) . in order to determine which model applies to abell 3112 , temperature and abundance measurements of the excess x - ray components would be useful because the excess components are expected to have higher abundances and lower entropy in the entrainment scenario .
unfortunately , the present data have too few counts in these regions to allow a detailed spectral analysis .
a deeper radio image would also be useful , as would lower frequency observations to search for steep spectrum emission .
note that the effect of the central radio source is limited to a region very close ( @xmath76 ) to the center .
this region is much smaller than the cooling flow region itself ( @xmath77 ) . outside the interacting region
, there is no evidence of the interaction such as x - ray cavities , instead the x - ray image is quite symmetric and smooth .
therefore , the central radio source does not directly affect the structure of the cooling flow outside of the very central region , at least at present .
note that we only have high frequency radio data .
low frequency radio observations might give another picture .
in the case of m87 in the virgo cluster , for instance , 327 mhz radio data show an morphology which fills out a large portion of the central region of the cluster ( owen , eilek , & kassim 2000 ) , although similar structures can not be seen in higher frequency data .
radiative cooling produces a temperature decrease towards the center of the cluster . on the other hand , thermal conduction would reduce this temperature gradient .
if we adopt the standard spitzer conductivity , conduction would be expected to be very effective and to have eliminated the central temperature gradient ( figure [ fig : radtctcond ] ) .
however , the observations show that the temperature does decrease significantly at the center of the cluster .
this implies that thermal conduction is significantly suppressed below the spitzer value .
although we did not find a large amount of locally cooling multiphase gas , the existence of a large - scale temperature gradient indicates that the icm is cooling in a global scale .
the mass deposition rate of abell 3112 determined from our spectroscopy of the total spectrum is comparable to that determined previously based on _ rosat _ and _ exosat _ image analyses ( edge , stewart , & fabian 1992 ; allen & fabian 1997 ; peres et al .
however , the temperature range of the cooled gas is limited to @xmath0 kev .
in addition , emission measure distribution about temperature is probably different than that expected from the standard cooling flow model .
this suggests that some other heating mechanism may affect the cooling gas .
one probable solution is heat conduction which is reduced from the spitzer value by some physical mechanism but still is energetically important ( e.g. , hattori & umetsu 2000 ; malyshkin & kulsrud 2001 ) .
the relativistic electrons from the central radio source might also supply the required heating ( e.g. , fabian et al .
2000 ; mcnamara et al .
2000 ; david et al . 2001
; blanton et al . 2001 ) .
however , it is unlikely that the radio - emitting electrons are heating all of the cooling gas in abell 3112 at present , because the radio lobes are so small that they do not directly affect the cooling flow on large scales . in abell 3112
, there is no evidence for ghost bubbles which might indicate a history of past radio source interactions ( mcnamara et al .
another possible form of radio source heating would be high energy protons from the central agn .
since the minimum energy density in relativistic electrons is too small to support radio bubbles against the surrounding hot gas ( e.g. , blanton et al . 2001 ) , most of the energy in the radio source may be in the form of relativistic protons .
the accelerated protons could diffuse and heat the icm through coulomb interactions ( rephaeli & silk 1995 ; inoue & sasaki 2001 ) .
although they do not emit any observable radio radiation , protons would produce several hundred mev @xmath78-rays through the decay of neutral pions produced in collisions with thermal protons .
future @xmath78-ray observations will test this hypothesis .
we would like to thank h. bhringer , j. c. kempner , and y. fujita for very useful comments .
support for this work was provided by the national aeronautics and space administration through _ chandra _ award number go1 - 2133x issued by the _
chandra _ x - ray observatory center , which is operated by the smithsonian astrophysical observatory for and on behalf of nasa under contract nas8 - 39073 .
m. t. was supported in part by a grant - in - aid from the ministry of education , science , sports , and culture of japan ( 13440061 ) .
support for e. l. b. was provided by nasa through the _ chandra _ fellowship program , grant award number pf1 - 20017 , under nasa contract number nas8 - 39073 .
allen , s. w. , & fabian , a. c. 1997 , mnras , 286 , 583 allen , s. w. , schmidt , r. w. , & fabian , a. c. 2002 , mnras , 334 , l11 binney , j. , & tremaine , s. 1987 , galactic dynamics ( princeton : princeton univ . press ) blanton , e. l. , sarazin , c. l. , mcnamara , b. r. , & wise , m. w. 2001 , apj , 558 , l15 bhringer , h. , nulsen , p. e. j. , braun , r. , & fabian , a. c. 1995 , mnras , 274 , l67 churazov , e. , brggen , m. , kaiser , c. r. , bhringer , h. , & forman , w. 2001 , apj , 554 , 261 david , l. p. , jones , c. , & forman , w. 1995 , apj , 445 , 578 dickey , j. m. , & lockman , f. j. 1990 , ara&a , 28 , 215 edge , a. c. , stewart , g. c. , & fabian , a. c. 1992 , mnras , 258 , 177 ettori , s. , & fabian , a. c. 1999 , 305 , 834 fabian , a. c. 1994 , ar&aa , 32 , 277 fabian , a. c. , et al .
2000 , mnras , 318 , 65 feigelson , e. d. , wood , p. a. d. , schreier , e . j. , harris , d. e. , & reid , m. j. 1987 apj , 312 , 101 finoguenov , a. , david , l. p. , & ponman , t. j. 2000 , apj , 544 , 188 fujita , y. , sarazin , c. l. , kempner , j. c. , rudnick , l. , slee , o. b. , roy , a. l. , andernach , h. , & ehle , m. 2002 , apj , 575 , 764 hattori , m. , & umetsu , k. 2000 , apj , 533 , 84 inoue , s. , & sasaki , s. 2001 , apj , 562 , 618 irwin , j. a. , bregman , j. n. , & evrard , a. e. apj , 1999 , 519 , 518 jing , y. p. , & suto , y. 2002 , apj , 574 , 538 kaastra , j. s. 1992 , an x - ray spectral code for optically thin plasmas ( internal sron - leiden report , updated version 2.0 ) kaastra , j. s. , et al . , 2001 , a&a 365 , l99 lee , j. & suto , y. 2003 , apj , 585 , 151 liedahl , d. a. , osterheld , a. l. , & goldstein , w. h. 1995 , apj , 522 , 82 makishima , k. , et al . , 2001 ,
pasj , 53 , 401 malyshkin , l. , & kulsrud , r. 2001 , apj , 549 , 402 markevitch , m. , forman , w. r. , sarazin , c. l. , & vikhlinin , a. 1998 , apj , 503 , 77 mcnamara , b. r. & oconnell , r. , w. 1989 , aj , 98 , 2018 mcnamara , b. r. , et al .
2000 , apj , 534 , l135 mcnamara , b. r. , et al . 2001 ,
apj , 562 , l149 mohr , j. j. , mathiesen , b. , & evrard , a. e. 1999 , apj , 517 , 627 peres , c. b. , fabian , a. c. , edge , a. c. , allen , s. w. , johnstone , r. m. , & white , d. a. 1998 , mnras , 298 , 416 owen , f. n. , eilek , j. a. , & kassim , n. e. 2000 , apj , 543 , 611 peterson , j. r. , et al .
2001 , a&a , 365 , l104 peterson , j. r. , et al .
2003 , apj submitted , astro - ph/0210662 rephaeli , y. & silk , j. 1995 , apj , 442 , 91 sarazin , c. s. , 1986 , rev .
phys . , 58 , 1 spitzer , l. , jr .
1962 , physics of fully ionized gases ( new york : wiley ) tamura , t. , et al .
2001 , a&a , 365 , l87 cccccc + @xmath79 & @xmath80 & @xmath81 & @xmath73 & @xmath1 & @xmath82 + ( arcsec ) & ( kev ) & ( @xmath83 ) & ( @xmath84 @xmath18 ) & ( @xmath5 yr@xmath6 ) & + @xmath85 & @xmath86 & @xmath87 & @xmath88 & @xmath89 & 284.5/200 + @xmath90 & @xmath91 & @xmath92 & @xmath93 & @xmath94 & 285.3/233 + @xmath95 & @xmath96 & @xmath97 & @xmath98 & @xmath99 & 226.2/216 + @xmath100 & @xmath101 & @xmath102 & @xmath103 & @xmath104 & 234.3/205 + @xmath105 & @xmath106 & @xmath107 & @xmath108 & @xmath109 & 222.9/195 + @xmath110 & @xmath111 & @xmath112 & @xmath113 & @xmath114 & 216.8/188 + @xmath115 & @xmath116 & @xmath117 & @xmath118 & @xmath119 & 202.5/181 + @xmath120 & @xmath121 & @xmath122 & @xmath123 & @xmath124 & 204.5/175 + cccccc + @xmath79 & @xmath80 & @xmath81 & @xmath73 & @xmath1 & @xmath82 + ( arcsec ) & ( kev ) & ( @xmath83 ) & ( @xmath84 @xmath18 ) & ( @xmath5 yr@xmath6 ) & + @xmath85 & @xmath125 & @xmath126 & ( @xmath127 ) & @xmath128 & 315.6/201 + @xmath90 & @xmath129 & @xmath130 & ( @xmath127 ) & @xmath131 & 302.5/234 + @xmath95 & @xmath132 & @xmath133 & ( @xmath127 ) & @xmath134 & 237.1/217 + @xmath100 & @xmath135 & @xmath136 & ( @xmath127 ) & @xmath137 & 236.9/206 + @xmath105 & @xmath138 & @xmath139 & ( @xmath127 ) & @xmath140 & 230.1/196 + @xmath110 & @xmath141 & @xmath142 & ( @xmath127 ) & @xmath143 & 217.7/189 + @xmath115 & @xmath144 & @xmath145 & ( @xmath127 ) & @xmath146 & 202.5/182 + @xmath120 & @xmath147 & @xmath148 & ( @xmath127 ) & @xmath149 & 205.85/176 + cccccc + @xmath80 & @xmath150 & @xmath81 & @xmath73 & @xmath1 & @xmath82 + ( kev ) & ( kev ) & ( @xmath83 ) & ( @xmath84 @xmath18 ) & ( @xmath5 yr@xmath6 ) & + @xmath151 & @xmath152 & @xmath153 & @xmath154 & @xmath155 & @xmath156 + @xmath157 & @xmath158 & @xmath159 & @xmath160 & @xmath161 & @xmath162 + cccccc + @xmath163 & @xmath81 & @xmath73 & @xmath164 & @xmath165 & @xmath82 + ( kev ) & ( @xmath83 ) & ( @xmath84 @xmath18 ) & & ( @xmath84 @xmath18 ) & + @xmath166 & @xmath167 & @xmath168 & @xmath169 & @xmath170 & 12.9/21 + @xmath171 & @xmath172 & @xmath160 & @xmath173 & @xmath174 & 12.9/21 + note .
the abundance and @xmath73 of the mekal component were derived from the fit of the region surrounding the point source . upper and lower rows show the results when the @xmath73 is allowed to vary or fixed to the galactic value , respectively . | we present the results of a _
chandra _ observation of the central region of abell 3112 .
this cluster has a powerful radio source in the center and was believed to have a strong cooling flow .
the x - ray image shows that the intracluster medium ( icm ) is distributed smoothly on large scales , but has significant deviations from a simple concentric elliptical isophotal model near the center .
regions of excess emission appear to surround two lobe - like radio - emitting regions .
this structure probably indicates that hot x - ray gas and radio lobes are interacting . from an analysis of the x - ray spectra in annuli
, we found clear evidence for a temperature decrease and abundance increase toward the center .
the x - ray spectrum of the central region is consistent with a single - temperature thermal plasma model .
the contribution of x - ray emission from a multiphase cooling flow component with gas cooling to very low temperatures locally is limited to less than 10% of the total emission .
however , the whole cluster spectrum indicates that the icm is cooling significantly as a whole , but in only a limited temperature range ( @xmath0 kev ) .
inside the cooling radius , the conduction timescales based on the spitzer conductivity are shorter than the cooling timescales .
we detect an x - ray point source in the cluster center which is coincident with the optical nucleus of the central cd galaxy and the core of the associated radio source .
the x - ray spectrum of the central point source can be fit by a 1.3 kev thermal plasma and a power - law component whose photon index is 1.9 .
the thermal component is probably plasma associated with the cd galaxy .
we attribute the power - law component to the central agn . |
demixing of binary or multicomponent mixtures is a very common phenomenon observed in a broad range of molecular fluids @xcite , polymer solutions and blends @xcite , or colloidal dispersions @xcite .
phase separation is generally associated with differences in the attractive interactions between particles of different chemical species . in polymer solutions these differences
are usually embodied in the flory @xmath0-parameter @xcite , which controls the competition between the entropy of mixing and the total interaction energy , at least within a mean - field picture . on the other hand , in multi - component colloidal systems like binary dispersions involving colloidal particles of very different sizes , or mixtures of colloidal particles and non - adsorbing
polymer , phase separation can be driven by purely repulsive , excluded volume interactions . by mapping the initial multi - component system onto
an _ effective _ one - component system involving only the bigger colloidal particles , the largely entropy - driven demixing can be understood in terms of _ attractive _ depletion interactions induced between the large particles by the smaller species ( the `` depletant '' ) @xcite
. it should however be kept in mind that the initial bare interactions are purely repulsive , albeit strongly non - additive , as in the highly simplified asakura - oosawa model for colloid - polymer mixtures @xcite . in the case of fully additive hard sphere mixtures ,
phase separation has been predicted for sufficiently large size ratios @xcite , but it is now generally believed that the fluid - fluid demixing is metastable , and preempted by freezing @xcite . a significant degree of positive non - additivity of the core radii @xmath1 ( whereby @xmath2 , with @xmath3 ) is required to observe a stable demixing transition in the fluid phase @xcite .
effective interactions between the centres of mass of fractal objects , like linear polymer coils @xcite , star polymers @xcite or dendrimers @xcite , obtained by averaging over individual monomer degrees of freedom , are now known to be very `` soft '' .
more specifically the effective pair potential diverges only logarithmically for overlapping star polymers @xcite , while remaining finite , of the order of @xmath4 , for linear polymers in good solvent @xcite .
this observation has stimulated the investigation of simple models , like finite repulsive step potentials @xcite , or the gaussian core potential @xcite , which was first introduced by stillinger , in a somewhat different context @xcite , namely @xmath5 where @xmath6 is the energy scale , while @xmath7 determines the range of the effective potential .
it is worth stressing that the `` gaussian core '' model is unrelated to the `` gaussian molecule '' model , which was extensively studied by michael fisher and collaborators @xcite . in the latter model
it is the mayer @xmath8-function , rather than the pair potential , which has a gaussian shape .
simple , penetrable particle models are also widely used in highly coarse - grained simulations of large - scale phenomena within the so - called `` dissipative particle dynamics '' ( dpd ) method @xcite . in dpd
, effective interactions between penetrable fluid `` particles '' are frequently modelled by a simple parabolic potential @xcite : @xmath9 it has recently been realized that binary mixtures of particles with penetrable cores , which interact via generalizations of the gaussian and parabolic potentials and , involving different energy scales @xmath6 and radii @xmath7 for the various species , may phase - separate over appropriate ranges of these parameters .
spinodal instability was first shown to occur for the gaussian core model within the random phase approximation ( rpa ) @xcite and binodals as well as interfacial properties were then calculated within the same approximation @xcite .
similarly gibbs ensemble monte carlo simulations have very recently shown that binary systems of soft particles interacting via the parabolic potential phase separate beyond a critical degree of enhanced repulsion between particles of different species , in agreement with flory - like mean field considerations @xcite . in this paper
we systematically extend our earlier results for the gaussian core model @xcite and investigate the range of validity of the rpa by detailed calculations of the pair structure , thermodynamics , and the resulting phase coexistence curve within the much more accurate hypernetted chain ( hnc ) approximation .
the break - down of the rpa is quantified in the physically relevant regime where @xmath10 and @xmath11 , which would correspond to the cross - over from dilute to semi - dilute regimes of the underlying binary polymer solution .
the rpa continues to provide reliable first estimates at higher densities .
the model under consideration is the binary gaussian core model ( gcm ) already introduced in references @xcite and @xcite .
it consists of @xmath12 particles of `` radius '' @xmath13 and @xmath14 particles of `` radius '' @xmath15 in a volume @xmath16 .
the total number density is @xmath17 , while the concentrations of the two species are @xmath18 and @xmath19 , respectively .
the pair potentials are @xmath20 which introduce three length and three energy parameters : @xmath21 , @xmath22 , @xmath23 and @xmath24 .
if @xmath21 is chosen as unit of length , the system is entirely specified by the 5 dimensionless parameters @xmath25 , @xmath26 , @xmath27 , @xmath28 , @xmath29 , where @xmath30 . for a fixed set of dimensionless parameters
the reduced helmholtz free energy per particle , @xmath31 , is a function only of the intensive variables @xmath32 and @xmath33 ; this may be split into the ideal gas , ideal mixing and excess ( non - ideal ) contributions : [ eq:4 ] @xmath34 where @xmath35 is an irrelevant de broglie thermal wavelength .
the equation of state @xmath36 , and the chemical potentials @xmath37 follow from the standard thermodynamic relations : [ eq:5 ] @xmath38 and may likewise be split into ideal and excess contributions . within rpa and hnc
, these excess contributions may be easily expressed in terms of the usual total and direct correlation functions @xmath39 and @xmath40 , which are related by the familiar ornstein - zernike ( oz ) relations @xcite ; these may be used to express the fourier transforms ( ft ) @xmath41 of the total correlation functions in terms of the ft @xmath42 of the direct correlation functions : [ eq:6 ] @xmath43,\\ \hat{h}_{12}(k ) & = \frac{1}{\delta(k ) } \hat{c}_{12}(k),\\ \hat{h}_{22}(k ) & = \frac{1}{\delta(k ) } \left[\hat{c}_{22}(k)(1-\rho_1 \hat{c}_{11}(k ) ) + \rho_1 \hat{c}_{12}^2(k)\right ] , \end{aligned}\ ] ] where @xmath44 and @xmath45[1-\rho_2 \hat{c}_{22}(k ) ] - \rho_1 \rho_2 \hat{c}_{12}^2(k).\ ] ] these oz relations must be supplemented by closure relations .
the rpa amounts to simply identifying the @xmath40 with their asymptotic ( large @xmath46 ) behaviour : @xmath47 substitution of the ft s of into the relations and yields explicit expressions for the @xmath41 , which may be transformed back to obtain the @xmath48 .
detailed analytical expressions are given in appendix [ sec : solut - viri - route ] in the special case of mixtures with equal core radii @xmath49 .
the hnc closure relations are @xcite : [ eq:9 ] @xmath50 an iterative procedure must be used to solve the coupled equations and numerically .
convergence is easily achieved for the binary gcm model over most of the potential parameter space .
in fact the hnc approximation becomes exact in the high density limit ( @xmath51 ) , and is extremely accurate for densities @xmath52 , as shown earlier in the one - component case @xcite .
pair structure data will be examined in greater detail in section 5 .
knowledge of the pair correlation functions allows direct access to thermodynamics , via the compressibility or virial routes @xcite .
being approximate , the rpa and hnc closures are not thermodynamically consistent , i. e. the two routes lead to different answers . in the case of the gcm
the two theories become strictly thermodynamically consistent only in the high density limit , @xcite . within the rpa ,
the compressibility route is equivalent to a simple mean - field ansatz for the free energy @xcite .
this leads immediately to the following analytic expression for the excess part of the reduced free energy @xcite : [ eq:10 ] @xmath53 where @xmath54 . in the special case of the gcm : @xmath55 the pressure and chemical potentials @xmath56 follow directly from eqns .
( the superscript c refers to the compressibility route ) : [ eq:12 ] @xmath57 the virial route is considerably more arduous .
the pressure follows from the virial theorem : @xmath58 in the case of the symmetric mixture ( @xmath59 ; @xmath60 ) the rpa pair correlation functions obtained in appendix a may be used to yield the following expression ( the superscript v referring to the virial route ) : @xmath61 where the functions @xmath62 and the roots @xmath63 and @xmath64 are defined in appendix [ sec : solut - viri - route ] ( eqns . and ) . the resulting reduced free energy per particle is obtained by integrating the equation of state with respect to density : @xmath65,\ ] ] where the polylogarithm @xmath66 is defined in equation .
unfortunately this expression becomes singular for densities larger than the spinodal densities obtained by the compressibility route .
this mathematical artifact discussed in appendix [ sec : solut - viri - route ] does not allow us to use expressions and to construct a phase diagram .
the lack of thermodynamic consistency of the rpa is illustrated in figure [ fig:1 ] , which compares the compressibility and virial equations of state as a function of density , for a symmetric mixture ; the virial pressures are always lower than their compressibility counterparts , and closer to the nearly exact hnc results , to which we now turn .
the hnc approximation is very nearly thermodynamically consistent in practice , at least for temperatures and densities relevant for polymer solutions @xcite .
hence we have calculated thermodynamic properties within the more convenient virial route with the pressure given by the standard relation , while the chemical potentials may also be directly expressed in terms of the direct and total correlation functions , according to @xcite : @xmath67 - \rho_\mu \hat{c}_{\nu \mu}(k=0)\right\}.\end{aligned}\ ] ] note that these expressions hold only within the hnc approximation , and are consistent with the virial route .
as mentioned earlier , suitably reduced equilibrium properties of the binary gcm depend on the five dimensionless combinations @xmath68 , @xmath69 , @xmath70 , @xmath71 , and @xmath72 . the compressibility version of the rpa thermodynamics ( or equivalently , the mean field approximation ) allows a considerable reduction of this parameter space . from the expressions - it is clear that the radii @xmath1 and energies @xmath73 enter only in the combinations @xmath74 defined in eqn . .
in terms of the reduced density @xmath75 and pressure @xmath76 , the thermodynamic behaviour of the mixture is uniquely determined by the dimensionless ratios @xmath77 and @xmath78 .
this remarkable reduction of parameter space holds only for the thermodynamics , but not for the correlation functions , which depend explicitly on the five parameters of the gcm potentials .
the scale invariance of the thermodynamics is broken by the rpa virial route , as is immediately evident form the explicit expressions and ( valid for a symmetric mixture ) ; the same is true within the hnc approximation .
however deviations from rpa - c ( mean - field ) scale invariance are expected to be small in the high density , high temperature regime , where rpa is increasingly accurate , so that the considerable reduction in the number of relevant potential parameters ( from 5 to 2 ) is expected to carry over , at least approximately , to the phase diagrams calculated with rpa - v or hnc thermodynamics , which will be presented in the next section .
the phase behaviour of the binary gcm may be deduced from the knowledge of the reduced free energy per particle @xmath8 as a function of the variables @xmath79 and @xmath33 and from its thermodynamic derivatives ( the pressure and the chemical potentials ) . for suitable values of the potential parameters @xmath73 and @xmath80 ,
the binary gcm becomes unstable against demixing at sufficiently high densities . in terms of the free energy per particle , @xmath81
, the standard thermodynamic stability conditions of a binary mixture read @xcite : [ eq:17 ] @xmath82 the first condition ensures mechanical stability , while the second inequality guarantees stability against demixing at constant volume ; the third inequality ensures stability of the mixture at constant pressure .
note that the stability at constant pressure is the more restrictive condition .
the subsequent discussion is restricted to demixing at constant pressure .
the vanishing of the l. h. s. of equation , corresponding to the case where the third inequality turns into an equality , signals the occurrence of spinodal instability . within the mean - field ( rpa - c ) approximation , the equation for the spinodal
is easily calculated to be @xcite @xmath83 where : [ eq:19 ] @xmath84 it is easily inferred from that a spinodal instability occurs whenever @xmath85 .
this region is visualized in figure [ fig:2 ] in a plot of the minimum energy ratio @xmath86 required for spinodal instability , as a function of the size ratio @xmath87 .
the radius @xmath88 is taken to be given by the combination rule : @xmath89 as suggested by renormalization group ( rg ) calculations @xcite and by direct simulation @xcite of mixtures of self - avoiding polymer coils .
figure [ fig:2 ] also shows the values of the ratio @xmath90 , as calculated for two self - avoiding walk polymers by rg techniques @xcite , plotted as a function of the ratio of their radii of gyration .
the figure suggests that phase separation of real polymers would only be observed for ratios of gyration radii larger than 10 , i. e. for mixtures of very long and very short polymers .
however , the rg results are probably not trustworthy for such asymmetric mixtures , and the rg potentials are strictly valid only in the infinite dilution limit . the spinodal line in the @xmath91-plane is easily calculated from to be : @xmath92 ^ 2 + 4 x ( 1-x ) \delta v}}{2 x ( 1-x ) \delta v}\ ] ] the critical concentration is determined by the condition : @xmath93 where @xmath94 is the pressure calculated from , at the spinodal density given by .
the corresponding critical density @xmath95 , calculated by substituting the critical concentration determined by is plotted in figure [ fig:3 ] versus the ratio @xmath78 , for several ratios @xmath96 .
the binodal or phase coexistence curves in the @xmath97 plane are determined by the usual conditions of equality of the chemical potentials of both species and of the pressures in the two phases , using expressions - .
the resulting equations must in general be solved numerically to yield the binodal curves .
the complete phase diagram can however be calculated analytically within the rpa - c approximation , in the symmetric case , where @xmath60 . due to the symmetry of the problem
, all thermodynamic quantities must be invariant with respect to the transformation @xmath98 and @xmath99 .
this simplification allows the fully analytic treatment detailed in appendix b. the resulting spinodal and binodal are shown in figure [ fig:4 ] in the case where @xmath100 .
the same figure also shows the hnc binodals , for several values of @xmath70 ( remember that within rpa - c all these binodals coincide ) .
as expected the hnc results break the scale invariance of rpa - c and shift the binodal curves upward , to higher densities .
the hnc and rpa - c critical densities differ by almost a factor of 2 for @xmath101 , pointing to the limitations of the mean - field ( rpa - c ) description .
an example of a phase diagram in an asymmetric mixture , with potential parameters already used in earlier rpa calculations @xcite , is shown in figure [ fig:5a ] in the @xmath97 plane and in figure [ fig:5b ] in the @xmath91 plane .
the spinodal and binodal curves are now asymmetric , and the difference between rpa - c and hnc results is again significant , with the hnc binodal being pushed to higher densities .
phase separation in binary mixtures of particles interacting via the closely related parabolic core potentials has been observed in the gibbs ensemble monte carlo simulations of wijmans et al .
they considered the symmetric case where all @xmath1 are equal , @xmath102 and @xmath103 .
we have calculated the binodal in the @xmath104 plane under the same conditions as the mc simulations , i. e. for @xmath105 , in the rpa and hnc approximations .
the results are compared to the mc data in figure [ fig:6a ] . in view of the fact that the amplitude of the parabolic repulsion is @xmath106 , which makes it more hard - core like
, the agreement may be considered to be rather satisfactory . as in the gcm case ,
the hnc coexistence - curve lies well above its rpa counterpart ; it is closer to the simulation data .
the pair correlation functions @xmath48 of the gcm mixture can be calculated by combining the oz relations with either the rpa closure or the hnc closure . in the symmetric case , characterized by a single gaussian range parameter @xmath59 , analytic expressions , in the form of infinite series ,
can be obtained for the rpa closure , as shown in appendix a. in the asymmetric case , solutions for the @xmath48 are readily obtained by numerical fourier transformation of equations and , using the straightforward fourier transforms of the @xmath40 given by .
the hnc closure requires an iterative solution of the coupled oz and closure equations .
this was achieved using the standard picard method , and well - converged solutions were generally obtained in a few iterations .
convergence was found to be slower in the vicinity of phase coexistence , and to break down rapidly inside the phase coexistence region corresponding to metastable mixtures .
concentration fluctuations build up in that region , as signalled by a @xmath107 peak of growing amplitude in the @xmath41 , or equivalently in the corresponding partial structure factors @xmath108 the concentration - concentration structure factor : @xmath109 satisfies the long wavelength limit @xcite : @xmath110 where @xmath111 is the gibbs free energy .
the cross - over from metastable to unstable mixture corresponds to the vanishing of @xmath112 , so that @xmath113 is expected to diverge along a spinodal line .
this is indeed the case with the rpa closure , but the hnc closure ceases to converge before a spinodal line is reached , a well - known shortcoming of the hnc closure @xcite . such a break - down is typical of the inadequacy of fluid integral equations to describe critical behaviour @xcite .
the accuracy of hnc in the stable one - phase region for the pair correlation function @xmath114 , already documented in the one - component gcm @xcite , is tested under more severe conditions ( namely @xmath115 ) in figure [ fig:6 ] against recent monte carlo data @xcite for the parabolic core potential .
the agreement is seen to be excellent .
this suggests that the hnc closure would be very useful to generate pair structure in coarse - grained dpd fluids at a moderate computational cost .
examples of pair correlation functions and partial structure factors for several states of gcm mixtures are shown in figures [ fig:7]-[fig:9 ] .
the most striking feature is the amount of structure observed both in @xmath46 and in @xmath116-space , even within the rpa , compared to the previously studied one - component case @xcite .
the peak in the structure factors at @xmath117 as the spinodal is approached is expected , as explained earlier .
more surprising perhaps is the appearance of peaks in the pair correlation functions as @xmath118 , or at finite @xmath46 .
peaks at @xmath119 are of course precluded in the presence of hard cores , but are quite significant in the present penetrable core model and point to a novel physical mechanism to trigger phase separation , namely the clustering of the minority species , as observed e. g. in figure [ fig:9 ] for the symmetric case .
the clustering can be explained by a simple `` energetic '' argument : the particles of the majority species tend to maximise their mutual distances .
if the repulsion between unlike particles is stronger than between like ones , particles of the minority species prefer to cluster in the voids left by the majority species , rather than overlap with particles of the latter .
this mechanism drives phase separation .
the two - component extension of the gaussian core model leads to non - trivial phase behaviour , which we have investigated within rpa and hnc theories .
the former is partially analytic , while the latter requires only a modest numerical effort to calculate partial pair correlation functions and the resulting thermodynamic properties . while the pair structure shows some unexpected features , the most interesting prediction is the occurrence of phase separation induced by purely repulsive pair interactions .
as expected for this model of penetrable particles , phase separation is driven by an enhanced repulsion between unlike particles compared to the like - particle interaction .
a special feature of the gcm mixtures is that at high densities of both species , the mixture behaves like a `` mean field fluid '' , i. e. the rpa becomes asymptotically exact @xcite .
this means that at finite densities , the rpa makes semi - quantitavely valid predictions for the phase behaviour , and allows a rapid exploration of potential parameter space to search for likely conditions for phase separation .
the nature of the link between the behaviour of the binary gcm and that of a binary polymer solution must still be worked out in detail , mainly because the effective interactions between the centres of mass of the polymer coils are state - dependent @xcite . for a one - component system at lower densities , the structure is not very sensitive to the state - dependence @xcite .
we therefore expect the binary gcm to make qualitatively correct predictions for the structure of binary polymer solutions in the dilute and beginning of the semi - dilute regimes .
the link to the phase - behaviour is less clear .
because the gcm uses a fixed potential , it is mainly relevant to behaviour in the dilute regime .
while polymers in a melt are known to phase - separate rather easily @xcite , in the dilute regime for good solvents they are not expected to phase - separate @xcite .
this appears to be confirmed by the results shown in fig 2 .
however , it may be that for poorer solvent , or for stronger polymer incompatibilities , a regime where the polymers phase - separate at low densities could open up .
if this is the case , then the binary gcm would be a useful coarse - grained model with which one could rapidly explore the qualitative phase behaviour of this regime .
another open question relates to the critical behaviour of the binary gcm in the immediate vicinity of the critical consolute point .
the rpa clearly predicts mean field exponents.as regards the exact exponents , it is not obvious whether the correct critical behaviour of the model necessarily belongs to the ising universality class , since there is no clear correspondence between gcm mixtures and standard lattice models , due to the penetrability of the particles . the question of the correct universality class , a subject dear to the heart of michael fisher , must be considered as open .
r. finken is indebted to the oppenheimer fund for their support .
a. a. louis would like to thank the isaac newton trust , cambridge , for funding .
we consider the gcm where the potentials have equal range @xmath120 , and the heights of the repulsion potentials of like particles are the same for both species @xmath121 . without loss of generality we can set @xmath122 .
then we have @xmath123 with @xmath124 if we introduce the abbreviation @xmath125 and set @xmath126 , the oz equations reduce to ( using equation ):
[ eq : b3 ] @xmath127 the denominator @xmath128 in equation ( [ eq:6 ] ) has the zeros [ eq : b4 ] @xmath129 the positive root @xmath63 is the spinodal density at any given composition .
the virial equation gives [ eq : b5 ] @xmath130 evaluating this integral in fourier space and making use of the parseval theorem leads to @xmath131 e^{-r^2/r^2 } \hat{h}_{\mu \nu}(k)\nonumber\\ & = \frac{1}{12 } \int d\mathbf{k } \frac{\hat{z}(k)^2}{\delta(k ) } [ x_1 x_2 \hat{z}(k ) \delta v - 2 x_1 x_2 \delta v - 1 ] [ 6 - k^2 r^2 ] \end{aligned}\ ] ] this integral can not be evaluated analytically in general .
however , if the density is below the spinodal density for a given composition , one can expand the denominator in the integrand in a power series with respect to @xmath132 .
the integral can then be evaluated term by term , interchanging the integration and the summation . in order to make the symmetry between the two species explicit
, we write @xmath133 .
the shifted concentration @xmath134 now varies between @xmath135 and 1 , with @xmath136 corresponding to @xmath137 . with the definition of the polylogarithmic function @xmath138 and the auxiliary function @xmath139\ ] ] evaluation of leads to @xmath140\\ & -\frac{z_2 ( 1-\xi^2)\delta v}{4 \pi^{3/2 } r^3 } \aleph\left(-\frac{\rho}{z_1}\right)\\ & -\frac{z_1 ( 1-\xi^2)\delta v}{4 \pi^{3/2 } r^3 } \aleph\left(-\frac{\rho}{z_2}\right ) .
\end{split}\ ] ] this can be integrated to give the free energy per particle @xmath141\\ & + \frac{1}{2 \pi^{3/2 } \rho r^3}[\operatorname{li}_{5/2}(\rho / z_1 ) + \operatorname{li}_{5/2}(\rho / z_2 ) + \rho ] \end{split}\ ] ] from that we get the chemical potentials as @xmath142 \nonumber\\ & + \frac{1}{2 \pi^{3/2 } r^3 ( 1-\xi ) \rho } \operatorname{li}_{3/2}\left(\frac{\rho}{z_2}\right ) \left[1 - \frac { \xi}{\sqrt{1 + \delta v ( 1-\xi^2)}}\right],\\ \beta \mu_2 & = \log(\lambda^3 \rho ) + \log\left(\frac{1+\xi}{2}\right ) + \frac{1+\xi}{2 } \rho + v_{12 } \frac{1- \xi}{2 } \rho + \frac{1}{2 \pi^{3/2 } r^3}\nonumber\\ & + \frac{1}{2 \pi^{3/2 } r^3 ( 1+\xi ) \rho } \operatorname{li}_{3/2}\left(\frac{\rho}{z_1}\right ) \left[1 - \frac { \xi}{\sqrt{1 + \delta v ( 1-\xi^2)}}\right ] \nonumber\\ & + \frac{1}{2 \pi^{3/2 } r^3 ( 1+\xi ) \rho } \operatorname{li}_{3/2}\left(\frac{\rho}{z_2}\right ) \left[1 + \frac { \xi}{\sqrt{1 + \delta v ( 1-\xi^2)}}\right].\end{aligned}\ ] ] it is easily checked that @xmath143 , due to the symmetry of the model .
note that these expressions become singular when @xmath144 .
this is a mathematical artefact originating in the interchange of summation and integration in , which renders the expressions for @xmath145 and @xmath146 unusable for binodal calculations .
the phase diagram simplifies considerably in the symmetric case where @xmath60 .
we express @xmath147 as @xmath148 and note that demixing occurs when the dimensionless parameter @xmath149 is positive . as above we introduce the reduced quantities @xmath150 and @xmath151 . because of the symmetry of the problem , all quantities must be invariant with respect to the transformation @xmath152 , which corresponds to a relabeling of species 1 and 2 . as in appendix
[ sec : solut - viri - route ] , we make use of the shifted concentration @xmath134 to make this symmetry explicit .
the pressure and the chemical potentials can now be expressed in terms of the variables @xmath153 and @xmath134 : two coexisting phases @xmath156 must have the same pressure and chemical potentials , respectively : @xmath157 because of the symmetry in this model , coexisting phases must have the same density @xmath158 and the corresponding concentrations must fulfill @xmath159 . considering this ,
equal pressure in both phases is trivially fulfilled , and the two conditions of equal chemical potentials become equivalent .
the relation between the binodal density @xmath160 and concentration is @xmath161 expanding the binodal densities around the critical point gives @xmath162.\ ] ] from the series expansions we see that for all @xmath163 the spinodal lies inside the binodal . within the rpa approximation
we thus get an analytically exact solution for a fluid - fluid demixing transition of a liquid consisting of purely repulsive particles . as expected in this treatment ,
the resulting critical exponents are mean - field exponents .
van duijneveldt , a.w .
heinen , and h.n.w .
* 21 * , 369 ( 1993 ) ; a. d. dinsmore , a. g. yodh , and d. j. pine , phys . rev .
e * 52 * , 4045 ( 1995 ) ; a. imhof and j. k. g. dhont , phys .
rev . lett . * 75 * , 1662 ( 1995 ) . | a two - component system of penetrable particles interacting via a gaussian core potential is considered , which may serve as a crude model for binary polymer solutions .
the pair structure and thermodynamic properties are calculated within the random phase approximation ( rpa ) and the hypernetted chain ( hnc ) integral equation .
the analytical rpa predictions are in semi - quantitative agreement with the numerical solutions of the hnc approximation , which itself is very accurate for gaussian core systems .
a fluid - fluid phase separation is predicted to occur for a broad range of potential parameters .
the pair structure exhibits a nontrivial clustering behaviour of the minority component .
similiar conclusions hold for the related model of parabolic core mixtures , which is frequently used in dissipative particle dynamics ( dpd ) simulations . |
we consider a problem of optimal debt management in infinite time horizon , modeled as a noncooperative game between a borrower and a pool of risk - neutral lenders .
since the debtor may go bankrupt , lenders charge a higher interest rate to offset the possible loss of part of their investment . in the models studied in @xcite ,
the borrower has a fixed income , but large values of the debt determine a bankruptcy risk .
namely , if at a given time @xmath2 the debt - to - income ratio @xmath0 is too big , there is a positive probability that panic spreads among investors and bankruptcy occurs within a short interval @xmath3 $ ] .
this event is similar to a bank run .
calling @xmath4 the random bankruptcy time , this means @xmath5\,\big|\,t_b
> t\big\}=\rho(x(t))\cdot\varepsilon+ o(\varepsilon).\ ] ] here the `` instantaneous bankruptcy risk '' @xmath6 is a given , nondecreasing function . at all times
@xmath2 , the borrower must allocate a portion @xmath7 $ ] of his income to service the debt , i.e. , paying back the principal together with the running interest .
our analysis will be mainly focused on the existence and properties of an optimal repayment strategy @xmath8 in feedback form . in the alternative model proposed by nuo and thomas in @xcite
, the yearly income @xmath9 is modeled as a stochastic process : @xmath10 here @xmath11 is an exponential growth rate , while @xmath12 denotes brownian motion on a filtered probability space . differently from @xcite , in @xcite it is the borrower himself that chooses when to declare bankruptcy .
this decision will be taken when the debt - to - income ratio reaches a certain threshold @xmath1 , beyond which the burden of servicing the debt becomes worse than the cost of bankruptcy . at the time @xmath13 when bankruptcy occurs , we assume that the borrower pays a fixed price @xmath14 , while lenders recover a fraction @xmath15 $ ] of their outstanding capital . here
@xmath16 is a nondecreasing function of the debt size .
for example , the borrower may hold an amount @xmath17 of collateral ( gold reserves , real estate@xmath18 ) which will be proportionally divided among creditors if bankruptcy occurs . in this case , when bankruptcy occurs each investor will receive a fraction @xmath19 of his outstanding capital .
aim of the present paper is to provide a detailed mathematical analysis of some models closely related to @xcite .
we stress that these problems are very different from a standard problem of optimal control .
indeed , the interest rate charged by lenders is not given a priori .
rather , it is determined by the expected evolution of the debt at all future times .
hence it depends globally on the entire feedback control @xmath20 .
a `` solution '' must be understood as a nash equilibrium , where the strategy implemented by the borrower represents the best reply to the strategy adopted by the lenders , and conversely .
our main results can be summarized as follows .
* we first assume that value @xmath1 at which bankruptcy occurs is a priori given , and seek an optimal feedback control @xmath8 which minimizes the expected cost to the borrower . for any value @xmath21 of the diffusion coefficient in , we prove that the problem admits at least one nash equilibrium solution , in feedback form . in the deterministic case
where @xmath22 , the solution can be constructed by concatenating solutions of a system of two odes , with terminal data given at @xmath23 . *
* we then study how the expected total cost of servicing the debt together with the bankruptcy cost ( exponentially discounted in time ) , are affected by different choices of @xmath1 .
let @xmath24 $ ] be the salvage rate , i.e. , the fraction of outstanding capital that will be payed back to lenders if bankruptcy occurs when the debt - to - income ratio is @xmath1 .
if @xmath25 then , letting @xmath26 , the total expected cost to the borrower goes to zero . on the other hand ,
if @xmath27 then the total expected cost to the borrower remains uniformly positive as @xmath26 .
we remark that the assumption is quite realistic .
for example , if holds , then @xmath28 for all @xmath1 large enough .
we remark that rules out the possibility of a _ ponzi scheme _
, where the old debt is serviced by initiating more and more new loans . indeed , if holds , then such a strategy will cause the total debt to blow up to infinity in finite time .
the remainder of the paper is organized as follows . in section 2
we describe more carefully the model , deriving the equations satisfied by the value function @xmath29 and the discounted bond price @xmath30 . in sections 3 and 4
we construct equilibrium solutions in feedback form , in the stochastic case ( @xmath31 ) and in the deterministic case ( @xmath22 ) , respectively .
finally , sections 5 and 6 contain an analysis of how the expected cost to the borrower changes , depending on the bankruptcy threshold @xmath1 .
in the economics literature , some related models of debt and bankruptcy can be found in @xcite . a general introduction to nash equilibria and
differential games can be found in @xcite .
for the basic theory of optimal control and viscosity solutions of hamilton - jacobi equations we refer to @xcite .
we consider a slight variant of the model in @xcite .
we denote by @xmath32 the total debt of a borrower ( a government , or a private company ) at time @xmath2 .
the annual income @xmath9 of the borrower is assumed to be a random process , governed by the stochastic evolution equation .
the debt is financed by issuing bonds . when an investor buys a bond of unit nominal value , he receives a continuous stream of payments with intensity @xmath33 . here
* @xmath34 is the interest rate payed on bonds , which we assume coincides with the discount rate , * @xmath35 is the rate at which the borrower pays back the principal .
if no bankruptcy occurs , the payoff for an investor will thus be @xmath36 in case of bankruptcy , a lender recovers only a fraction @xmath37 $ ] of his outstanding capital . here
@xmath38 can depend on the total amount of debt at the time on bankruptcy . to offset this possible loss
, the investor buys a bond with unit nominal value at a discounted price @xmath39 $ ] . as in @xcite , at any time @xmath2 the value @xmath40 is uniquely determined by the competition of a pool of risk - neutral lenders .
we call @xmath41 the rate of payments that the borrower chooses to make to his creditors , at time @xmath2 .
if this amount is not enough to cover the running interest and pay back part of the principal , new bonds are issued , at the discounted price @xmath40 .
the nominal value of the outstanding debt thus evolves according to @xmath42 the debt - to - income ratio is defined as @xmath43 . in view of and ( [ dx ] ) , ito s formula @xcite yields the stochastic evolution equation @xmath44\ , dt - \sigma \,x(t)\ , dw.\ ] ] here @xmath45 is the portion of the total income allocated to pay for the debt . throughout the following we assume @xmath46 .
we observe that , if @xmath47 , then the borrower s income grows faster than the debt ( even if no payment is ever made ) . in this case , with probability one the debt - to - income ratio would approach zero as @xmath48 .
n this model , the borrower has two controls . at each time
@xmath2 he can decide the portion @xmath49 of the total income which he allocates to repay the debt .
moreover , he can decide at what time @xmath13 bankruptcy is declared . throughout the following , we consider a control in feedback form , so that @xmath50,\ ] ] while bankruptcy is declared as soon as @xmath0 reaches the value @xmath1 .
the bankruptcy time is thus the random variable @xmath51 at first , we let @xmath1 be a given upper bound to the size of the debt . in a later section , we shall regard @xmath1 as an additional control parameter , chosen by the borrower in order to minimize his expected cost .
given an initial size @xmath52 of the debt , the total expected cost to the borrower , exponentially discounted in time , is thus computed as @xmath53~=~e\left[\int_0^{t_b } e^{-rt } l(u^*(x(t)))\ , dt + e^{-r t_b } b\right]_{x(0)=x_0}.\ ] ] here @xmath14 is a large constant , accounting for the bankruptcy cost , while @xmath54 is the instantaneous cost for the borrower to implement the control @xmath55 . in the following
we shall assume * ( a ) * _ the cost function @xmath56 is twice continuously differentiable for @xmath57 and satisfies @xmath58 _ for example , one may take @xmath59 for some @xmath60 .
fix @xmath61 . for a given initial debt
@xmath62 $ ] , we define the corresponding value function as @xmath63\,.\ ] ] under the assumptions * ( a ) * we have @xmath64 denote by @xmath65 } % { 0\leq \omega\leq \min\{(\lambda+r)x,1\ } } ~\left\ { l(\omega ) - \dfrac{\xi}{p}\ , \omega\right\}+\left(\dfrac{\lambda+r}{p } -\lambda + \sigma^2 -\mu \right ) x\ , \xi\,\ ] ] the hamiltonian associated to the dynamics ( [ db ] ) and the cost function @xmath56 in ( [ scost ] ) . notice that , as long as @xmath66 , the function @xmath67 is differentiable with lipschitz continuous derivatives w.r.t .
all arguments . by a standard arguments
, the value function @xmath29 provides a solution to the second order ode @xmath68 with boundary conditions ( [ bd4 ] ) .
as soon as the function @xmath29 is determined , the optimal feedback control is recovered by @xmath69}{\mathrm{argmin}}\left\ { l(\omega ) - \dfrac{v'(x)}{p(x)}\ , \omega\right\}.\ ] ] by * ( a ) * this yields @xmath70 % \min\left\{(l')^{-1}\big(\dfrac{v'(x)}{p(x)}\big),(\lambda+r)x\right\}\qquad\qquad \displaystyle ( l')^{-1}\left(\dfrac{v'(x)}{p(x)}\right)&\qquad \mathrm{if}\qquad \displaystyle\dfrac{v'(x)}{p(x)}~>~l'(0)\,.\end{cases}\ ] ] assuming that lenders are risk - neutral , the discounted bond price @xmath30 is determined by @xmath71_{x(0)=x_0}\,,\ ] ] where @xmath38 denotes the salvage rate . in other words , if bankruptcy occurs when the debt - to - income ratio is @xmath1 , then investors receive a fraction @xmath24 $ ] of the nominal value of their holding .
notice that the random variable @xmath13 in ( [ tb ] ) now depends on the initial state @xmath52 , the threshold @xmath1 , and on the feedback control @xmath72 . by the feynman - kac formula
, @xmath73 satisfies the equation @xmath74\cdot p'(x)+ \dfrac{(\sigma x)^2}{2 } p''(x),\ ] ] with boundary values @xmath75 combining ( [ eqv ] ) and ( [ phj ] ) , we are thus led to the system of second order odes @xmath76 ( r+\lambda)(p(x)-1)&=~\displaystyle h_{\xi}\bigl(x , v'(x),p(x)\bigr)\cdot p'(x)+\dfrac{(\sigma x)^2}{2}\cdot p''(x)\ , , \end{cases}\ ] ] with the boundary conditions @xmath77 we close this section by collecting some useful properties of the hamiltonian function . [ lemma : hprop ] let the assumptions * ( a ) * hold .
then , for all @xmath78 and @xmath39 $ ] , the function @xmath67 in ( [ h1 ] ) satisfies @xmath79 @xmath80 moreover , for every @xmath81 the map @xmath82 is concave down and satisfies @xmath83 + \infty,&\textrm { if } ~~\dfrac{1}{p}\leq\left(\dfrac{\lambda+r}{p } -\lambda + \sigma^2 -\mu\right ) x\ , .
\end{cases}\end{aligned}\ ] ] 1 . 2 .
since @xmath84 is defined as the infimum of a family of affine functions , it is concave down .
we observe that ( [ h1 ] ) implies @xmath85 this yields the identities ( [ h00])-([h0x ] ) .
3 . 4 . taking @xmath86 in ( [ h1 ] ) we obtain the upper bound in ( [ hb1 ] ) . by the concavity property ,
the map @xmath87 is non - increasing . hence ( [ h0x ] ) yields the upper bound in ( [ hb2 ] ) .
6 . since @xmath88 for all
@xmath89 $ ] , we have @xmath90}\left\{-\dfrac{\xi}{p}w\right\}+\left(\dfrac{\lambda+r}{p } -\lambda + \sigma^2 -\mu \right ) x\ , \xi\,\ ] ] and obtain the lower bound in ( [ hb1 ] ) . on the other hand , using the optimality condition , one computes from ( [ h1 ] ) that @xmath91 where @xmath92}{\mathrm{argmin}}~\left\ { l(\omega ) - \dfrac{\xi}{p}\ , \omega\right\}~=~(l')^{-1}\left(\dfrac{\xi}{p}\right)~<~1\,.\ ] ] observe that , as @xmath93 , one has @xmath94 in ( [ hxx ] ) . the non - increasing property of the map @xmath95 yields the lower bound in ( [ hb2 ] ) . 7 .
8 . to prove ( [ limh ] ) we observe that , in the first case , there exists @xmath96 such that @xmath97 hence , letting @xmath93 we obtain @xmath98~=~-\infty\,.\ ] ] to handle the second case , we observe that , for @xmath99 large , the minimum in ( [ h1 ] ) is attained at the unique point @xmath100 where @xmath101 . hence @xmath102 and @xmath103\\[4 mm ] & \displaystyle\geq~\lim_{\xi\to + \infty } l(\omega(\xi))~=~+\infty.\end{aligned}\ ] ]
let @xmath61 be given . if a solution @xmath104 to the boundary value problem ( [ ode0])-([bdc ] ) is found , then the feedback control @xmath8 defined at ( [ u^ * ] ) and the function @xmath105 provide an equilibrium solution to the debt management problem . in other words , for every initial value @xmath52 of the debt , the following holds . * consider the stochastic dynamics ( [ db ] ) , with @xmath106 and @xmath107
. then for every @xmath108 $ ] the identity ( [ p ] ) holds , where @xmath13 is the random bankruptcy time defined at ( [ tb ] ) .
* given the discounted price @xmath105 , for every initial data @xmath109 $ ] the feedback control @xmath8 is optimal for the stochastic optimization problem @xmath110.\ ] ] with stochastic dynamics ( [ db ] ) , where @xmath107 . to construct a solution to the system ( [ ode0])-([bdc ] ) , we consider the auxiliary parabolic system @xmath111 p_t(t , x)&=~(r+\lambda)(1-p(t , x))+ \displaystyle h_{\xi}\bigl(x , v_x(t , x),p(t , x)\bigr)\cdot p_x(t , x)+\dfrac{(\sigma x)^2}{2}\cdot p_{xx}(t , x)\ , , \end{cases}\ ] ] with boundary conditions ( [ bdc ] ) . following @xcite ,
the main idea is to construct a compact , convex set of functions @xmath112\mapsto [ 0,b]\times [ \theta(x^*),1]$ ] which is positively invariant for the parabolic evolution problem . a topological technique
will then yield the existence of a steady state , i.e. a solution to ( [ ode0])-([bdc ] ) .
[ thm : existence ] in addition to * ( a ) * , assume that @xmath31 and @xmath113 .
then the system of second order odes ( [ ode0 ] ) with boundary conditions ( [ bdc ] ) admits a @xmath114 solution @xmath115 , such that @xmath116\to [ 0,b]$ ] is increasing and @xmath117\to [ \theta(x^*),1]$ ] is decreasing .
if @xmath22 , then the stochastic equation ( [ db ] ) reduces to the deterministic control system @xmath195 we then consider the deterministic debt management problem . *
_ given an initial value @xmath62 $ ] of the debt , minimize @xmath196 subject to the dynamics ( [ detode ] ) , where the bankruptcy time @xmath13 is defined as in ( [ tb ] ) , while @xmath197 _ since in this case the optimal feedback control @xmath198 and the corresponding functions @xmath193 may not be smooth , a concept of equilibrium solution should be more carefully defined .
[ def : eqsol ] a couple of piecewise lipschitz continuous functions @xmath8 and @xmath199 provide an equilibrium solution to the debt management problem ( dmp ) , with continuous value function @xmath200 , if * for every @xmath108 $ ] , @xmath200 is the minimum cost for the optimal control problem @xmath201 subject to @xmath202 moreover , every carathodory solution of ( [ cont1 ] ) with @xmath203 is optimal . * for every @xmath108 $ ] , there exists at least one solution @xmath204 of the cauchy problem @xmath205 such that @xmath206 with @xmath13 as in ( [ tb ] ) . in the deterministic case , ( [ ode0 ] )
takes the form @xmath207 ( r+\lambda)(p(x)-1)&=~h_\xi\bigl(x , v'(x),p(x)\bigr)\,p'(x)\,,\end{cases}\ ] ] with hamiltonian function ( see fig . [
f : g53 ] ) @xmath208}~\left\ { l(\omega ) - \frac{\xi}{p } \omega \right\}+\left(\dfrac{\lambda+r}{p } -(\lambda + \mu)\ , \right ) x\ , \xi\,.\ ] ] we consider solutions to ( [ ode1 ] ) with the boundary condition @xmath209 introduce the function @xmath210 with the understanding that @xmath211 if @xmath212 , recalling ( [ detode ] ) we define @xmath213 notice that @xmath214 is the control that keeps the debt @xmath144 constant in time .
this value @xmath214 achieves the minimum in ( [ h2 ] ) when @xmath215 this motivates the definition @xmath216 on the other hand , if @xmath217 , then the function @xmath82 is monotone increasing and we define @xmath218 .
observe that @xmath219 h_{\xi}(x,\xi , p)&<~0\qquad\forall \xi>\xi^{\sharp}(x , p)\ , .
\end{cases}\ ] ] we regard the first equation in ( [ ode1 ] ) as an implicit ode for the function @xmath29 . for every @xmath220 and @xmath39 $ ] , if @xmath221 , then this equation has no solution . on the other hand , when @xmath222 the implicit ode ( [ ode1 ] ) can equivalently be written as a differential inclusion ( fig .
[ f : g53 ] ) : @xmath223 , the the hamiltonian function @xmath82 has a global maximum @xmath224 . for @xmath225 , the values @xmath226 are well defined .
, title="fig : " ] [ f : g53 ] recalling ( [ detode ] ) , we observe that * the value @xmath227 corresponds to the choice of an optimal control such that @xmath228 .
* the value @xmath229 corresponds to the choice of an optimal control such that @xmath230 .
* when @xmath231 , then the value @xmath232 corresponds to the unique control such that @xmath233 .
since @xmath82 is concave down , the functions @xmath234 satisfy the following monotonicity properties ( fig .
[ f : g53 ] ) * _ for any fixed @xmath147 , the map @xmath235 is decreasing , while @xmath236 is increasing . _ for @xmath237 , the second ode in ( [ ode1 ] ) can be written as @xmath238 where @xmath239 consider the function @xmath240 with the understanding that @xmath241 if @xmath242 .
notice that @xmath243 is the total cost of keeping the debt constantly equal to @xmath144 ( in which case there would be no bankruptcy and hence @xmath244 ) .
moreover , denote by @xmath245 the solution to the system of odes @xmath246 p'(x)&=~g^{-}(x , v(x),p(x))\ , , \end{cases}\ ] ] with terminal conditions @xmath247 next , consider the point @xmath248\,;~~v_b(x)<w(x)\big\},\ ] ] and call @xmath249 the solution to the backward cauchy problem @xmath250,\\[3 mm ] v(x_1)&=~w(x_1 ) .
\end{cases}\ ] ] we will show that a feedback equilibrium solution to the debt management problem is obtained as follows ( see fig . [
f : g55 ] ) .
@xmath251,\\ & \\
v_b(x)\qquad \hbox{if}\quad x\in [ x_1 , x^*].\end{cases}\label{sol1}\\ p^*(x)&=\begin{cases}1,&\textrm { if } x\in [ 0 , x_1],\\ & \\ p_b(x),&\textrm { if } x\in \,]x_1 , x^*].\end{cases}\label{sol2}\\ u^*(x)&=\begin{cases}\displaystyle\underset{\omega\in [ 0,1]}{\mathrm{argmin } } \left\{l(\omega ) - \frac{(v^*)'(x)}{p^*(x)}\omega\right\},&\textrm { if } x\ne x_1,\\ & \\ ( r-\mu)x_1,&\textrm { if } x= x_1\,.\end{cases}\label{sol3}\end{aligned}\ ] ] , the debt increases until it reaches @xmath252 , then it is held at the constant value @xmath252 .
if the initial debt is @xmath253 , the debt keeps increasing until it reaches bankruptcy in finite time.,title="fig : " ] [ f : g55 ] [ t:41 ] assume that the cost function @xmath56 satisfies the assumptions * ( a ) * , and moreover @xmath254 .
then the functions @xmath255 in ( [ sol1])([sol3 ] ) are well defined , and provide an equilibrium solution to the debt management problem , in feedback form . 1 . 2 .
the solution of ( [ ode2])-([td2 ] ) satisfies the obvious bounds @xmath256.\ ] ] we begin by proving that the function @xmath257 is well defined and strictly positive for @xmath258x_1 , x^*]$ ] .
+ to prove that @xmath259x_1,x^*]\,,\ ] ] assume , on the contrary , that @xmath260 for some @xmath261 . by the properties of the function
@xmath262 ( see fig . [
f : g53 ] ) it follows @xmath263 for some constant @xmath264 and all @xmath265 $ ] , @xmath266 $ ] .
hence , for any solution of ( [ ode2 ] ) , @xmath267 implies @xmath268 for all @xmath269 , providing a contradiction .
+ next , observe that the functions @xmath270 are locally lipschitz continuous as long as @xmath271 . moreover , @xmath272 implies @xmath273 therefore , the functions @xmath274 are well defined on the interval @xmath275 $ ] .
3 . 4 .
if @xmath276 the construction of the functions @xmath255 is already completed in step * 1.*. in the case where @xmath277 , we claim that the function @xmath278 is well defined and satisfies @xmath279 indeed , if @xmath280 for some @xmath281 , the lipschitz property ( [ flip ] ) again implies that @xmath282 for all @xmath269 .
this contradicts the terminal condition in ( [ v1ode ] ) .
+ to complete the proof of our claim ( [ v1prop ] ) , it suffices to show that @xmath2830 , x_1].\ ] ] this is true because @xmath284 & \displaystyle=~f^-\bigl(x , h^{max}(x,1 ) , 1\bigr)~=~f^-(x , w(x),1 ) .
\end{aligned}\ ] ] 5 . 6 . in the remaining steps ,
we show that @xmath255 provide an equilibrium solution .
namely , they satisfy the properties ( i)-(ii ) in definition [ def : eqsol ] .
+ to prove ( i ) , call @xmath285 the value function for the optimal control problem ( [ cost1])-([cont1 ] ) .
+ for any initial value , @xmath286 , in both cases @xmath287 $ ] and @xmath288 x_1 , x^*]$ ] , the feedback control @xmath198 in ( [ sol3 ] ) yields the cost @xmath289 .
this implies @xmath290 to prove the converse inequality we need to show that , for any measurable control @xmath291 $ ] , calling @xmath204 the solution to @xmath292 one has @xmath293 where @xmath294 is the bankruptcy time ( possibly with @xmath295 ) . + for @xmath296 $ ] , consider the absolutely continuous function @xmath297 at any lebesgue point @xmath2 of @xmath20 , we compute @xmath298\\[3 mm ] & = \displaystyle ~e^{-rt}\left[l(u(t))-rv^*(x(t ) ) + ( v^*)'(x(t))\left(\left(\frac{\lambda+r}{p^*(x(t))}-\lambda-\mu\right)x(t)-\frac{u(t)}{p^*(x(t))}\right)\right]\\[3 mm ] & \displaystyle\ge~ e^{-rt}\left [ \min_{\omega\in[0,1 ] } \left\{l(\omega)-\frac{(v^*)'(x(t))}{p^*(x(t))}\,\omega\right\ } + \left(\frac{\lambda+r}{p^*(x(t))}-\lambda-\mu\right ) x(t)(v^*)'(x(t))-rv^*(x(t))\right]\\[3 mm ] & = ~e^{-rt}\big[h\left(x(t),(v^*)'(x(t ) ) , p^*(x(t))\right)-rv^*(x(t))\big]~=~ 0.\end{aligned}\ ] ] therefore , @xmath299 proving ( [ upvx_1 ] ) .
7 . 8 .
it remains to check ( ii ) .
the case @xmath300 is trivial .
two main cases will be considered .
+ _ case 1 : @xmath2880,x_1]$ ] .
_ then there exists a solution @xmath204 of ( [ fode ] ) such that @xmath301 and @xmath3020,x_1]$ ] for all @xmath303 .
moreover , @xmath304 in this case , @xmath305 and ( [ pb= ] ) holds .
_ case 2 : @xmath306x_1,x^*]$ ] .
_ then @xmath307 for all @xmath308 $ ] .
this implies @xmath309 from the second equation in ( [ ode1 ] ) it follows @xmath310 @xmath311 therefore , for every @xmath312 $ ] one has @xmath313 letting @xmath314 we obtain @xmath315 proving ( [ pb= ] ) .
the construction described at ( [ sol1])([sol3 ] ) uniquely determines a feedback equilibrium solution to the debt management problem . in general
, however , we can not rule out the possibility that a second solution exists .
indeed , if the solution @xmath274 of ( [ ode2])-([td2 ] ) can be prolonged backwards to the entire interval @xmath188 $ ] , then we could replace ( [ sol1])-([sol2 ] ) simply by @xmath316 , @xmath317 for all @xmath318 $ ] .
this would yield a second solution .
we claim that no other solutions can exist .
this is based on the fact that the graphs of @xmath319 and @xmath257 can not have any other intersection , in addition to @xmath320 and @xmath321 .
indeed , assume on the contrary that @xmath322 for some @xmath323 ( see fig .
[ f : g56 ] ) . since @xmath324 and @xmath325
, the inequalities @xmath326 yield a contradiction . next , let @xmath327 be any equilibrium solution and define @xmath328\,;~~p(x)=1\bigr\}.\ ] ] then * on @xmath329x^\dagger , x^*]$ ] the functions @xmath330 provide a solution to the backward cauchy problem ( [ ode2])-([td2 ] ) . * on @xmath3290 , x^\dagger]$ ]
the function @xmath331 provides the value function for the optimal control problem @xmath332 subject to the dynamics ( with @xmath244 ) @xmath333 and the state constraint @xmath334 $ ] for all @xmath138 .
the above implies @xmath335,\\[3 mm ] v^\dagger(x)~&\leq ~w(x ) , \textrm { if } x\in [ 0 , x^\dagger ] .
\end{cases}\ ] ] since @xmath331 must be continuous at the point @xmath336 , by the previous analysis this is possible only if @xmath337 or @xmath338 . in ( [ h2 ] )
, the graphs of @xmath257 and @xmath12 can not have a second intersection at a point @xmath339.,title="fig : " ] [ f : g56 ]
in this section we study the behavior of the value function @xmath257 when the maximum size @xmath1 of the debt , at which bankruptcy is declared , becomes very large . for a given @xmath61 ,
we denote by @xmath340 , @xmath341 the solution to the system ( [ ode2 ] ) with terminal data ( [ td2 ] ) . letting @xmath342
, we wish to understand whether the value function @xmath257 remains positive , or approaches zero uniformly on bounded sets . toward this goal
, we introduce the constant @xmath343 from ( [ ode1 ] ) and ( [ h2 ] ) it follows @xmath344 in turn , if @xmath345 , this implies @xmath346\,.\ ] ] calling @xmath8 the optimal feedback control , by ( [ u^ * ] ) we have @xmath347\,.\ ] ] in this case , the hamiltonian function takes a simpler form , namely @xmath348\cdot \frac{v'x}{p}\,,\\ h_{\xi}(x , v',p)&=~\bigl[(\lambda+r)-(\lambda+\mu)p\bigr]x\,.\end{aligned}\ ] ] therefore , the system of odes ( [ ode2 ] ) can be written as @xmath349x}\ , v \,,\\[6 mm ] p'&=~(\lambda+r)\cdot \dfrac{p(p-1)}{[(\lambda+r)-(\lambda+\mu)p]\ , x } \ , .
\end{cases}\ ] ] the second ode of in ( [ ode3 ] ) is equivalent to @xmath350 solving backward the above ode with the terminal data @xmath351 , we obtain @xmath352\,.\ ] ] therefore , @xmath353\,.\ ] ] different cases will be considered , depending on the properties of the function @xmath354 . by obvious modeling considerations , we shall always assume @xmath355,\qquad\qquad \theta'(x)\leq x\qquad\qquad\forall x\geq 0.\ ] ] we first study the case where @xmath38 has compact support . recall that @xmath356 is the constant in ( [ m1def ] ) .
assume that @xmath357 for some constant @xmath358 .
then , for any @xmath359 , the solution @xmath360 , @xmath341 of ( [ ode2])-([td2 ] ) satisfies @xmath361\,.\ ] ] by ( [ im - p1 ] ) and ( [ a - theta-1 ] ) , for every @xmath359 one has @xmath362\,.\ ] ] inserting this into the first ode in ( [ ode3 ] ) , we obtain @xmath363 in turn , this yields @xmath364 for all @xmath365 $ ] .
this means that bankruptcy instantly occurs if the debt reaches @xmath366 .
next , we now consider that case where @xmath367 for all @xmath144 .
@xmath368 if @xmath369 and @xmath370 , then @xmath371\,.\ ] ] in particular , for @xmath372 $ ] one has @xmath373 observe that @xmath374 is a strictly decreasing function of @xmath144 . for a fixed value of @xmath1 , let @xmath375 $ ] be the inverse function of @xmath341 . from ( [ ode3 ] ) ,
a direct computation yields @xmath376\,\chi(p)}\cdot v_b(\chi(p),x^*)\cdot \chi'(p ) \,,\\[6 mm ] \dfrac{d}{dp}p_b(\chi(p),x^*)&=(\lambda+r)\cdot \dfrac{p(p-1)}{[(\lambda+r)-(\lambda+\mu)p]\cdot \chi(p ) } \cdot \chi'(p)=1\ , .
\end{cases}\ ] ] from ( [ odep ] ) it follows @xmath377 solving the above ode with the terminal data @xmath378 , @xmath379 , we obtain @xmath380 hence @xmath381 recalling ( [ im - p1 ] ) , a direct computation yields ( [ ubv1 ] ) .
the upper and lower bounds for @xmath382 in ( [ bound - v ] ) now follow from ( [ lowbp ] ) and the inequality @xmath383 .
[ cor : infty - th ] assume that @xmath384 then the value function @xmath385 satisfies @xmath386 indeed , for @xmath387 we have @xmath388 , and ( [ vto0 ] ) follows from the second inequality in ( [ bound - v ] ) .
when @xmath389 , since the map @xmath390 is nondecreasing , we have @xmath391 [ cor : low - bv ] assume that @xmath392 then @xmath393 moreover , the followings holds . * if @xmath394 then @xmath395 * assume that there exist @xmath396 such that @xmath397 for all @xmath144 sufficiently large
. then , for each @xmath398 , there exists an optimal value @xmath399 such that @xmath400 it is clear that ( [ low - bv ] ) is a consequence of ( [ bound - v ] ) and ( [ b2 ] ) .
we only need to prove ( i ) and ( ii ) . for a fixed @xmath401
, we consider the functions of the variable @xmath1 alone : @xmath402 using ( [ ubv1 ] ) and ( [ im - p1 ] ) , we obtain @xmath403 \right)\,,\ ] ] and @xmath404 this implies @xmath405~=~\left[\dfrac{1}{1+\frac{r-\mu}{r+\lambda}\cdot\frac{q(x^*)}{1-q(x^*)}}-1\right]\cdot\left[\frac{\theta'(x^*)}{\theta(x^*)}+\frac{1}{x^*}\right ] \\+\frac{r-\mu}{(r+\lambda ) \left(1+\frac{r-\mu}{r+\lambda}\cdot\frac{q(x^*)}{1-q(x^*)}\right)}\cdot \frac{\theta'(x^*)}{1-\theta(x^*)}\,.\end{gathered}\ ] ] if ( [ ca1 ] ) holds , then ( [ y ] ) and ( [ re1 ] ) imply @xmath406~\leq~0\qquad \hbox{for all } ~x^*>x\geq m_1.\ ] ] hence the function @xmath407 is non - increasing .
this proves ( [ inf - v ] ) .
o prove ( ii ) , we observe that @xmath408 hence ( [ ca2 ] ) and ( [ re1 ] ) imply @xmath409~>~0,\ ] ] for all @xmath1 sufficiently large . by ( [ y ] )
this yields @xmath410 for all @xmath1 large enough .
hence there exists some particular value @xmath411 where the function @xmath412 attains its global minimum .
this yields ( [ minv ] ) .
in this section we study how the value function depends on the bankruptcy threshold @xmath1 , in the stochastic case where @xmath413 .
extensions of corollaries [ cor : infty - th ] and [ cor : low - bv ] , will be proved , constructing upper and lower bounds for the solution @xmath414 , @xmath415 of the system ( [ ode0])-([bdc ] ) , in the form @xmath416 1 . 2 .
we begin by constructing a suitable pair of functions @xmath417 .
let @xmath418 be the solution to the backward cauchy problem @xmath419 ( r+\lambda)(p_1 - 1)&=~\displaystyle\big(\frac{\lambda+r}{p_1}+\sigma^2\big ) x p'_1\,,\end{cases } \qquad\qquad \begin{cases}\tilde{v}_1(x^*)&= ~b , \\[6 mm ] p_1(x^*)&=~\theta(x^*).\end{cases}\ ] ] this solution satisfies @xmath420 @xmath421 using ( [ pv1 ] ) and ( [ p1s ] ) one obtains @xmath422 & \displaystyle=~p'_1(x ) \cdot\left(\frac{x}{p_1(x)}+\frac{\sigma^2+r+\lambda}{r+\lambda}\cdot \frac{1-\theta(x^*)}{(\theta(x^*)x^*)^{\frac{r+\lambda}{r+\lambda+\sigma^2}}}\cdot \frac{x^{\frac{\sigma^2}{r+\lambda+\sigma^2}}}{p_1(x)^{\frac{\lambda+r}{\lambda+r+\sigma^2}}}\right)\,.\end{aligned}\ ] ] since @xmath423 is monotone decreasing , it follows that @xmath424 for all @xmath2580 , x^*[\,$ ] .
in turn , this yields @xmath425 recalling ( [ hb2 ] ) , we have @xmath426 next , differentiating both sides of the first ode in ( [ pv1 ] ) , we obtain @xmath4270,x^*[\,.\ ] ] this implies @xmath4280,x^*[\,.\ ] ] recalling ( [ hb1 ] ) and ( [ pv1 ] ) , we obtain @xmath429 when @xmath430 , the map @xmath431 is monotone decreasing . defining @xmath432,\\[4 mm
] \tilde { v}(x)\qquad&\hbox{for}~~ x\in \big[\frac{1}{r+\lambda},x^*\big]\,,\end{cases}\ ] ] we thus have @xmath433 3 .
we now construct the functions @xmath434 .
defining @xmath435 a straightforward computation yields @xmath436 set @xmath437 and consider the continuous function @xmath438 for @xmath439 one has @xmath440 and hence @xmath441 on the other hand , for @xmath442x_2,x^*[$ ] and @xmath443 , one has @xmath444 , and @xmath445 recalling ( [ hb2 ] ) , we get @xmath446 & \leq~(r+\lambda)(1-p_2)+\big[\frac{(\lambda+r ) x-1}{p_2}+(\sigma^2-\lambda-\mu)x\big]p'_2(x)+\frac{\sigma^2x^2}{2}p_2''\\[4 mm ] & = ~(r+\lambda)(1-p_2)-\big[\frac{(\lambda+r ) x-1}{p_2}+(\sigma^2-\lambda-\mu)x\big]\cdot \frac{p_2(x)}{x}+\sigma^2p_2\\[4 mm ] & = ~(r+\lambda)(1-p_2)-\big[(\lambda+r)-\dfrac{1}{x}+(\sigma^2-\lambda-\mu)p_2\big]+\sigma^2p_2\\[4 mm ] & = ~\dfrac{1}{x}-(r-\mu)p_2~=~ -\frac{(r-\mu)\theta(x^*)x^*}{x}-\dfrac{1}{x}~<~0\,.\end{aligned}\ ] ] in particular , @xmath4470,x^ * [ , ~\xi\geq 0\,.\ ] ] + next , define @xmath448\,.\ ] ] for all @xmath449 $ ] , we thus have @xmath450 , and hence @xmath4510,1]\,.\ ] ] on the other hand , for @xmath258x_2 , x^ * ] $ ] we have @xmath452 recalling ( [ hb1 ] ) , ( [ p2 ] ) , ( [ sig1 ] ) and ( [ x*2 ] ) , we estimate @xmath453 & = b\cdot \big[rp_2-r+\big(\lambda+r-\dfrac{1}{x}+(\sigma^2-\lambda-\mu)p_2(x)\big)-\sigma^2p_2\big]\\[4 mm ] & = b\cdot \big(\lambda-\dfrac{1}{x}-(\lambda+\mu - r)p_2\big)~=~b\cdot \big[\lambda(1-p_2(x))+(r-\mu)p_2(x)-\frac{1}{x}\big]~>~0\end{aligned}\ ] ] for all @xmath258x_2,x^*[$ ] . + recalling ( [ x*2 ] ) , one has @xmath454x_2,x^*]\,.\ ] ] therefore the map @xmath455 is monotone decreasing on @xmath456 $ ] , for all @xmath457x_2,x^*]$ ] .
this implies @xmath458x_2,x^ * ] , ~~q\in ] 0,p_2(x)]\,.\ ] ] together with ( [ ine1 ] ) , we finally obtain @xmath4590,x^*[\ , , ~~q\in ] 0,p_2(x)]\,.\ ] ] relying on ( [ sp1 ] ) , ( [ sv1 ] ) , ( [ sp2 ] ) and ( [ v2-s ] ) , and using the same comparison argument as in the proof of theorem [ thm : existence ] we now prove [ thm:6.1 ] in addition to ( a1 ) , assume that @xmath31 and @xmath113 .
then the system ( [ ode0 ] ) with boundary conditions ( [ bdc ] ) admits a solution @xmath460 satisfying the bounds ( [ vpb ] ) for all @xmath140 $ ] . 1 . 2 . recalling @xmath152 in ( [ ddef ] )
, we claim that the domain @xmath461\times [ p_1(x),p_2(x)],\quad\forall x\in [ 0,x^ * ] \big\}\ ] ] is positively invariant for the semigroup @xmath462 , generated by the parabolic system ( [ parv])-([parp ] ) .
namely : @xmath463 indeed , from the proof of theorem [ thm : existence ] , we have @xmath4640,x^*[\,.\ ] ] we now observe that * for any @xmath465 with @xmath466 , by ( [ sp1 ] ) the function @xmath467 is a subsolution of the second equation in ( [ par1 ] ) .
similarly , by ( [ sp2 ] ) , the function @xmath468 is a supersolution .
* for any @xmath469 with @xmath39 $ ] and @xmath470 , by ( [ sv1-m ] ) the function @xmath471 is a supersolution of the first equation in ( [ par1 ] ) .
similarly , by ( [ v2-s ] ) , the function @xmath472 is a subsolution .
+ together , ( i)-(ii ) imply the positive invariance of the domain @xmath473 .
4 . using the same argument as in step * 4 *
of the proof of theorem [ thm : existence ] , we conclude that the system ( [ ode0])-([bdc ] ) admits a solution @xmath474 .
let the assumptions in theorem [ thm:6.1 ] hold .
if @xmath475 then , for all @xmath220 , the value function @xmath414 satisfies @xmath476 using ( [ p1 ] ) , ( [ v1 ] ) and theorem [ thm:6.1 ] , we have the estimate @xmath477 for all @xmath478 .
this implies that ( [ v00 ] ) holds for all @xmath478 .
since @xmath479 is monotone increasing , we then have @xmath480.\ ] ] this completes the proof of ( [ v00 ] ) .
let the assumptions in theorem [ thm:6.1 ] hold .
if @xmath481 then @xmath482 where the constants @xmath483 are defined as @xmath484 this follows from ( [ p2 ] ) , ( [ v2 ] ) and theorem [ thm:6.1 ] .
if we allow @xmath485 , then the equations ( [ ode1 ] ) admit the trivial solution @xmath486 , @xmath487 , for all @xmath220 .
this corresponds to a ponzi scheme , producing a debt whose size grows exponentially , without bounds . in practice
, this is not realistic because there is a maximum amount of liquidity that the market can supply .
it is interesting to understand what happens when this bankruptcy threshold @xmath1 is very large . * if @xmath488 , then it is convenient to choose @xmath1 as large as possible . by delaying the time of bankruptcy , the expected cost for the borrower , exponentially discounted in time , approaches zero . *
if @xmath489 and ( [ ca1 ] ) holds then it is still convenient to choose @xmath1 as large as possible .
however , by delaying the time of bankruptcy , the cost to the borrower remains uniformly positive . *
if @xmath490 and ( [ ca2 ] ) holds , then for every initial value @xmath52 of the debt there is a choice @xmath491 of the bankruptcy threshold which is optimal for the borrower .
examples corresponding to three cases ( i)(iii ) are obtained by taking @xmath492 with @xmath493 , @xmath494 , or @xmath495 , respectively .
t is important to observe that the choice of the optimal bankruptcy threshold @xmath1 is never time consistent " .
indeed , at the beginning of the game the borrower announces that he will declare bankruptcy when the debt reaches size @xmath1 .
based on this information , the lenders determine the discounted price of bonds .
however , when the time @xmath13 comes when @xmath496 , it is never convenient for the borrower to declare bankruptcy .
it is the creditors , or an external authority , that must enforce termination of the game . to see this ,
assume that at time @xmath13 when @xmath496 the borrower announces that he has changed his mind , and will declare bankruptcy only at the later time @xmath497 when the debt reaches @xmath498 .
if he chooses a control @xmath499 for @xmath500 , his discounted cost will be @xmath501 this new strategy is thus always convenient for the borrower . on the other hand
, it can be much worse for the lenders .
indeed , consider an investor having a unit amount of outstanding capital at time @xmath13 . if bankruptcy is declared at time @xmath13
, he will recover the amount @xmath502 .
however , if bankruptcy is declared at the later time @xmath497 , his discounted payoff will be @xmath503 to appreciate the difference , consider the deterministic case , assuming that @xmath354 is the function in ( [ th4 ] ) , with @xmath504 , and that @xmath505 . by the analysis at the beginning of section 5
, we have @xmath506 for all @xmath507 $ ] . replacing @xmath1 with @xmath508 in ( [ im - p1 ] )
we obtain that the solution to ( [ ode3 ] ) with terminal data @xmath509 satisfies @xmath510 if the investors had known in advance that bankruptcy is declared at @xmath511 ( rather than at @xmath23 ) , the bonds would have fetched a smaller price . in conclusion , if the bankruptcy threshold @xmath1 is chosen by the debtor , the only nash equilibrium can be @xmath485 . in this case
, the model still allows bankruptcy to occur , when total debt approaches infinity in finite time . | a problem of optimal debt management is modeled as a noncooperative game between a borrower and a pool of lenders , in infinite time horizon with exponential discount .
the yearly income of the borrower is governed by a stochastic process . when the debt - to - income ratio @xmath0 reaches a given size @xmath1 , bankruptcy instantly occurs .
the interest rate charged by the risk - neutral lenders is precisely determined in order to compensate for this possible loss of their investment . for a given bankruptcy threshold @xmath1 , existence and properties of optimal feedback strategies for the borrower
are studied , in a stochastic framework as well as in a limit deterministic setting .
the paper also analyzes how the expected total cost to the borrower changes , depending on different values of @xmath1 . |
light - induced 11-_cis _ @xmath1 all-_trans _ isomerization in retinal is a paradigmatic example of an important ultrafast photochemical reaction in biology @xcite .
the photoreaction is the first step in dim - light vision , and its high quantum yield , and formation of all-_trans _ product within 200 fs , contribute to the high efficiency of the phototransduction cascade @xcite .
the importance of retinal photoisomerization has made its mechanism the subject of numerous experimental and computational studies . a related biological process , _ cis - trans _ isomerization in retinoic acid , has become of increasing interest in its role in zebrafish hindbrain development@xcite . to study retinal dynamics ,
current pump - probe experiments use ultrashort femtosecond laser pulses to excite the retinal from the ground ( @xmath2 to the first excited ( @xmath3 ) electronic state .
the subsequent isomerization dynamics is then followed using a series of probe pulses , providing important insights into the sub-200 fs timescale of photoproduct formation @xcite , the coherent dynamics of all-_trans _ photoproduct @xcite , and most recently , the role of conical intersections in determining the reaction rate @xcite .
significantly , retinal photoisomerization occurs with high quantum yield ( @xmath4 ) , making the first step in visual phototransduction highly efficient @xcite .
measurements of the quantum yield @xmath5 as a function of excitation wavelength @xmath6 show a maximum at 500 nm and a small decline of @xmath75% as the laser wavelength increases from 500 nm to 570 nm @xcite . a detailed time - dependent computational study of retinal photoisomerization using a multilevel redfield theory @xcite based on a minimal two - state , two - mode ( 2d ) model of retinal chromophore@xcite provided insights into the role of avoided crossings , conical intersections , and dissipative dynamics in ultrafast energy conversion in biomolecules @xcite .
in addition , they qualitatively reproduce many salient features of isomerization dynamics of retinal in rhodopsin , including the high quantum yield . there and elsewhere
the quantum yield is defined dynamically , as the probability of forming the all-_trans _ product starting from the initial _ cis _ wavepacket created by an ultrafast franck - condon excitation from the ground _
cis _ state .
this time - dependent wavepacket view @xcite is well - suited to describe time - domain experiments on rhodopsin , in which ultrafast @xmath8 excitation creates a wavepacket on a highly excited state .
however , in nature , the excitation involves natural light .
such light has a very short coherence time ( 1.32 fs for sunlight ) compared to the fs laser pulses used experimentally @xcite , and is incident on the molecule for a far longer time . as a result ,
after some time , which is dependent on a number of conditions@xcite , the molecule is prepared in a mixture of stationary states that exhibits no coherent time evolution@xcite .
furthermore , these stationary states are , from a zeroth - order viewpoint , a linear superposition of _ cis _ and _ trans _ configurations . the existence and some unusual properties ( such as microsecond lifetimes ) of stationary eigenstates of this kind in large polyatomic molecules
were examined experimentally as early as 1977 @xcite .
the questions then arise as to how to extract and understand the quantum yield in the case of incoherent light excitation , and how to describe the _ cis - trans _ isomerization of retinal from a stationary eigenstate viewpoint that is appropriate for vision under natural light conditions .
a substantial step towards this major goal is provided in this paper . here
, we address this issue by first generalizing the concept of the quantum yield to the case of incoherent light excitation of a polyatomic molecule .
we note that in the limit of rapid decoherence of initial coherences in the reduced density matrix , the photoreaction quantum yield is stationary .
this definition is then applied to calculate the quantum yield of _ cis - trans _ photoisomerization of retinal using a minimal two - state , two - mode model@xcite .
we find excellent overall agreement between our calculated quantum yield and the previous time - dependent result of stock and hahn @xcite .
in addition , our calculations , in both the presence and absence of relaxation , agree with the observed value of the quantum yield at 500 nm and qualitatively reproduce its observed decline with increasing excitation wavelength @xcite .
these findings demonstrate the role of localization properties of the stationary states in determining the quantum yields , a crucial feature in the process induced with natural light .
note that there are various definitions of the quantum yield that are adopted in time - dependent studies .
below , we consistently utilize the parameters and approach of ref .
, which has become something of a `` standard model '' for basic retinal dynamics .
this imposes a number of consistency requirements that are discussed in sect .
[ summary ] . before proceeding further
, we emphasize that incoherent excitation of biomolecules embedded in condensed - phase environments involves two processes that occur simultaneously : ( a ) the creation of the stationary states by the incoherent light , and ( b ) the relaxation between the stationary states caused by the interaction with the environment . in this work ,
we treat these processes sequentially , assuming that the stationary states are formed first and then relax due to the system - environment coupling .
this allows for a better assessment of their individual roles .
the eigenstates formed on stage ( a ) are considered in secs .
ii and iii ; the effects of relaxation are considered in sec .
iv and , significantly , shown to have a negligible effect on the quantum yield of retinal photoisomerization treated within in the two - state , two - mode model @xcite .
the results , as seen below , motivate a future study of simultaneous incoherent excitation and relaxation in this and related model retinal systems .
we extract the quantum yield for the stationary case from the standard time dependent result , hence exposing their relationship .
reference defines the photoreaction quantum yield from the time - dependent view as @xmath9 where the time - dependent populations of 11-_cis _ and all-_trans _ isomers are defined as @xcite @xmath10 here @xmath11 is the reduced density matrix of the retinal subsystem , @xmath12 11-_cis _ or @xmath12 all-_trans _ , and @xmath13 are the projection operators , which divide the full range of the isomerization coordinate @xmath14 $ ] into the _ cis _ ( @xmath15 $ ] ) and _ trans _ regions ( @xmath16 $ ] ) , as illustrated in fig .
1 . in eq .
( [ populations ] ) , @xmath17 is the heaviside step function , and the operators @xmath18 project onto the ground and excited diabatic electronic states @xmath19 and @xmath20 .
qualitatively , @xmath21 can be thought of as projecting onto the adiabatic ground state reactant _ cis _ region and @xmath22 onto the adiabatic ground state product _ trans_. in this definition , the adiabatic excited state populations @xmath23 and @xmath24 are neglected assuming that these populations have decayed to zero in the long - time limit .
alternative definitions [ e.g. , including @xmath23 in the numerator and @xmath24 in the denominator of eq .
( [ y ] ) ] are certainly possible and will be explored in future work @xcite , and would necessitate a refitting of the potential parameters to the new definition of the time - dependent quantum yield .
as an example , consider impulsive fc excitation of retinal from its ground electronic and vibrational states to the first excited electronic state @xmath25 . here the density matrix of the system at time zero is given by @xcite @xmath26 where @xmath27 is the ground state with zero quanta in the torsional and coupling modes @xmath28 and @xmath29 .
it is convenient to express the density matrix in the system eigenstate basis defined by @xmath30 where @xmath31 and @xmath32 are time - independent eigenvalues and eigenfunctions . in this basis , with frequencies @xmath33 , @xmath34 .
hence the population dynamics is given by @xmath35 since the time - dependence arises from the oscillatory behavior of the coherences , ( i.e. , terms with @xmath36 ) , if @xmath37 then the state populations are time independent .
an important example of such a case is molecular excitation with incoherent light [ eq .
( [ p1t_trans ] ) ] , which populates , after some time , the eigenstates of @xmath38 without any subsequent coherent dynamics @xcite .
similarly , environmentally induced effects cause the loss of coherences . in this case , retaining only the diagonal terms in eq .
( [ p1t_trans ] ) , we have @xmath39 this expression may be made more physically transparent by noting that the diagonal elements of the reduced density matrix following impulsive fc excitation at @xmath40 correspond to the linear absorption spectrum of the molecule described by the spectral lineshape function @xcite @xmath41 , where @xmath42 is the excitation frequency measured from the ground vibrational state of the _ cis_-isomer with energy @xmath43 ( we assume that this is the only vibrational state populated prior to excitation )
. then it follows from eq .
( [ rho_t0 ] ) that @xmath44 where @xmath45 is the transition dipole moment operator , and we set @xmath46 a.u . without loss of generality [ since the @xmath47 prefactors multiplying state populations cancel out in the expression for @xmath48 in eq .
( [ y ] ) ] . therefore , eq . ( [ pdiag ] ) may be rewritten in the form @xmath49 hence , as expected , only the eigenvalues @xmath31 that have non - negligible probability of being excited by incoherent light make a contribution to the stationary populations in eq .
( [ pi ] ) . given eq .
( [ pi ] ) , the time - dependent definition of the quantum yield ( [ y ] ) extends to the frequency domain by defining @xmath50 with the `` pre - averaged '' populations @xmath51 given by eq .
( [ pi ] ) .
alternatively , we can first define a frequency - dependent _ trans _
/ _ cis _ probability ratio @xmath52 where @xmath53 are the expectation values of the projection operators in eq .
( [ pi ] ) . the resulting frequency - dependent quantum yield ( [ qy2omega ] ) can be averaged with the normalized spectral line shape function to give @xmath54 it is clear that the definitions ( [ qy1 ] ) and ( [ qy2 ] ) , termed the pre - averaged " and post - averaged " quantum yield , are not equivalent , but both provide a physically meaningful measure of ( frequency - dependent ) photoreaction efficiency .
specifically , the pre - averaged definition ( [ qy1 ] ) is consistent with the original definition of hahn and stock [ eq . ( [ y ] ) ] @xcite , whereas the post - averaged definition admits a clear physical interpretation as the degree of _ cis _ vs. _ trans _ character of a collection of eigenstates independently excited by incoherent light .
expressions related to eqs .
( [ qy1 ] ) and ( [ qy2 ] ) were previously obtained in references and , where the long - time limits of electronic state populations were associated with the phase volumes occupied by the wavepackets evolving on the upper and lower adiabatic potential energy surfaces
. equations ( [ p1t_trans ] ) , ( [ pi ] ) , ( [ qy1 ] ) , and ( [ qy2 ] ) form the central tool for the computations below .
they generalize the quantum yield in eq .
( [ y ] ) to the frequency domain relevant to stationary eigenstates , and provide a theoretical basis for the computational study of photoinduced isomerization in model retinal as described below .
the essential physics comes from the recognition that the established time dependent definition [ eq .
( [ y ] ) ] relies entirely on the projections onto domains of @xmath55 to define _ cis _ vs. _ trans _ configurations .
indeed , each of eqs .
( [ y ] ) , ( [ qy1 ] ) and ( [ qy2 ] ) have essentially the same meaning , the fraction of population in the ground adiabatic state that is , in the long time limit , in the _
trans_-configuration , disregarding any initial population in ground - state _ cis _ well .
this is achieved in eq .
( [ y ] ) by placing all population in an initial wavepacket on the excited electronic surface .
equations ( [ qy1 ] ) and ( [ qy2omega ] ) achieve this by putting all population initially into a stationary mixture on the excited potential surface .
what the latter two equations emphasize is that in the presence of incoherent light and decoherence , the appropriate states to consider in the long time limit are the stationary eigenstates of the hamiltonian .
they , as seen below , will allow stationary state insight into the efficiency of the isomerization process .
below , we apply this formulation to evaluate the quantum yield for the primary photoreaction in rhodopsin using the two - state two - mode model@xcite , which was previously applied @xcite to linear absorption @xcite , raman @xcite , and femtosecond pump - probe @xcite spectra of retinal in rhodopsin .
the two - state two - mode model and its multidimensional ( 25-mode ) extension @xcite have since been used to explore quantum dynamics and coherent control of _ cis - trans _ photoisomerization of retinal chromophore in rhodopsin @xcite .
we note that the model has the following advantages over the one - dimensional scenario ( e.g. ref . ):
( 1 ) the present model is two - dimensional , including a relevant bend degree of freedom ; ( 2 ) it accounts for the conical intersection between the ground and the first excited electronic states of retinal ; ( 3 ) it reproduces many salient features of the experimentally measured isomerization dynamics , including the ultrafast 200 fs timescale , the transient pump - probe spectra , and the energy storage of the photoreaction .
we emphasize that the two - state two - mode model and the associated definition of the quantum yield [ eq . ( [ y ] ) ]
were previously parametrized so as to reproduce various properties of the photoreaction@xcite of retinal in rhodopsin .
it neglects the effects of solvation , which are known to dramatically affect the photoisomerization dynamics in solution @xcite , but does include , due to the parameter fit , some features of the interaction of the two modes with the remaining molecular background .
our study is thus restricted to _ cis - trans _ photoisomerisation of retinal in rhodopsin .
the wavefunctions @xmath56 in eq . ( [ pi ] )
are the eigenfunctions of the 2d model hamiltonian given by @xcite @xmath57 where @xmath58 is the kinetic energy operator , @xmath55 is the tuning mode ( or generalized reaction coordinate ) corresponding to low - frequency torsional modes , and @xmath59 is the coupling mode that corresponds to high - frequency unreactive modes . the second term on the right - hand side of eq . ( [ model ] ) is the interaction potential in a basis spanned by the diabatic electronic functions @xmath60 with @xmath61 @xcite . a plot of the adiabatic pess obtained by diagonalizing the potential energy term in eq .
( [ model ] ) is shown in fig .
1 . the model parameters @xmath62 , @xmath63 , @xmath64 , @xmath65 , and @xmath6 , chosen to reproduce the femtosecond dynamics of retinal in rhodopsin @xcite , are ( in ev ) : @xmath66 , @xmath67 , @xmath68 , @xmath69 , @xmath70 , @xmath71 , @xmath72 , and @xmath73 . a total of 900 eigenvalues and eigenvectors of @xmath38 were calculated and used to assemble the matrix elements of the projector operators and the lineshape function in eqs .
( [ qy1 ] ) and ( [ qy2 ] ) .
the converged @xmath74 was in agreement with results in ref . .
interestingly , we found that the spectrum could be classified as integrable , insofar as the nearest neighbor distribution of energy levels @xmath75 shows a structure that can be fit with a poisson distribution @xmath76 with the local mean spacing @xmath77 @xmath78 ( see , e.g. , ref . ) .
such distributions are becoming of increasing interest due , e.g. to a recent proposal @xcite regarding chaos and transport in biological processes .
the upper panel of figure 1 shows torsional profiles of the adiabatic potential energy surfaces ( pes ) obtained by diagonalizing the hamiltonian matrix ( [ model ] ) at @xmath79 .
the pes profiles exhibit a conical intersection at @xmath80 and @xmath79 , clearly visible in the two - dimensional plot in the lower panel of fig 1 .
the _ cis _ isomer of retinal is localized in the potential well of the lower diabatic pes ( @xmath81 ) on the left - hand side .
photoexcitation by incoherent light ( represented by a green arrow ) populates a number of stationary eigenstates ( grey lines ) with mixed _
cis_-_trans _ character .
the quantum yield is determined by projecting these eigenstates onto their respective _ cis _ and _ trans _ regions of configuration space as discussed above .
figure 2 shows the frequency - dependent quantum yield @xmath82 given by eq .
( [ qy2omega ] ) as a function of excitation energy ( measured from the ground vibrational state of _
cis_-isomer ) . here
[ from eq .
( [ qy2omega ] ) ] , pure _ cis _ states correspond to @xmath83 and pure _ trans _ states correspond to @xmath84 ; thus , the magnitude of @xmath82 reflects the _ cis _ or _ trans _ character of a particular eigenstate with energy @xmath85 .
as expected , all molecular eigenstates that occur below the minimum energy of the _ trans _ well ( 11,000 @xmath78 , see fig .
2 ) have negligible quantum yields due to the absence of _ trans _ eigenstates in this low energy range .
above the 11,000 @xmath78 threshold , the quantum yield is a rapidly varying irregular function of @xmath86 , reflecting the strong mixing between the _ cis _ and _ trans _
components by the full hamiltonian .
such mixed _ cis - trans _ eigenstates are conceptually similar to the long - lived eigenstates of mixed singlet - triplet character observed in pentacene @xcite . interestingly , a closer inspection of fig .
2 , as shown in the figure insert , reveals the presence of purely _ cis _ ( or _ trans _ ) eigenstates with @xmath87 ( or 1 ) , which are qualitatively similar to the electronically localized eigenstates in vibronically coupled systems such as pyrazine@xcite .
the effect has recently been analyzed in the context of geometric phase - induced localization @xcite .
however , such `` extreme states '' may well be a result of low dimensionality of the model , as is indeed the case in pyrazine@xcite .
table i lists the values of the quantum yield computed using the stationary eigenstates of the 2d model ( the corresponding @xmath88 is shown by the green line in fig .
2 ) @xcite . both pre - averaged and
post - averaged results ( @xmath89 and @xmath90 ) agree extremely well with the previous time - dependent wavepacket result@xcite of @xmath91 .
the fact that the stationary quantum yields agree so well with time - dependent calculations @xcite makes clear how the quantum yield is directly manifest in the stationary eigenstates .
moreover , we also note agreement with the measured value @xcite of @xmath92 .
the agreement with experiment is a natural consequence of the two - state two - mode model @xcite being parametrized to reproduce the measured quantum yield in time - dependent calculations . in order to explore the effect of @xmath74 on the calculated quantum yields
, we replaced the 2d model lineshape function in eqs .
( [ qy1 ] ) and ( [ qy2 ] ) by the experimentally measured absorption profile of retinal in rhodopsin @xcite shown by the red line in fig .
2 . the result gives @xmath93 , in worse agreement with experiment than the value obtained with @xmath88 .
this confirms that the potential surfaces are optimized for behavior in the domain shown in green in fig . 2 , but would require further work to properly represent the _ cis / trans _ branching in other energy regions .
mathies and co - workers @xcite also measured @xmath94 observing a decline in the photoreaction efficiency below 500 nm . for comparison
, we calculate the wavelength - resolved stationary quantum yields by dividing the entire @xmath6 interval into 10 nm - wide bins , and average eqs .
( [ qy1 ] ) and ( [ qy2 ] ) over the eigenstates with energies falling into a particular bin . as with the overall quantum yield ,
two related , but not identical , definitions are possible . in the first , one calculates the averaged _ cis _ and _ trans _
populations @xmath95 where @xmath96 is the center frequency corresponding to @xmath97-th bin , and define : @xmath98 as an alternative , we directly average the frequency - dependent quantum yield ( [ qy2 ] ) over the entire bin @xmath99 where @xmath100 is a normalization factor that serves to correct for the change in absorption intensity due to the varying @xmath6 .
figure 3 compares the wavelength dependent quantum yield with the experimental results in the 500 - 570 nm range ( ref . ) . while the observed quantum yield declines monotonically with @xmath6 and stays constant below @xmath101 nm ,
our theoretical values oscillate over the entire range of @xmath6 . above 500 nm ,
the calculated quantum yields tend to decline with @xmath6 , in qualitative agreement with experiment ( the only exception being the value of @xmath102 at 540 nm ) . as expected from the above analysis
( see table i ) , switching lineshape functions has a dramatic effect on the calculated quantum yields .
in contrast with the results presented in table i , however , using the experimental @xmath74 improves the overall agreement with experiment , particularly at @xmath103 nm .
also shown in the inset of fig .
3 is the `` bare '' frequency - dependent quantum yield , @xmath104 here , the absence of the spectral lineshape function [ present in eqs .
( [ qy1 ] ) and ( [ qy2 ] ) ] reveals the variation in _ cis _
/ _ trans _ character of molecular eigenstates with zero oscillator strengths , which do not contribute to the physical quantum yield ( [ y2lambda ] ) .
the @xmath105 [ inset of fig .
3 ] is seen to generally decline with increasing @xmath6 .
this behavior reflects the appearance of _ trans _ states at energies above 11,000 @xmath78 , and that their density in this region is smaller than that of _ cis _ states ( see fig .
thus , the _ trans _ character of molecular eigenstates can be expected to decrease with decreasing @xmath86 , as is implicit in the experimental results . while the downward trend in the wavelength dependence of @xmath106 ( inset of fig .
3 ) is somewhat more pronounced than that observed for @xmath107 and @xmath102 ( fig .
3 ) , we note that the former can not be directly compared with experiment .
this is because the bare quantum yield [ eq . ( [ ybare ] ) ] does not include the variation of the spectral lineshape function @xmath74 with @xmath64 .
our computed time - independent definitions of the quantum yield [ eq . ( [ qy1 ] ) and ( [ qy2 ] ) ] assume that the eigenstate populations @xmath108 do not depend on time . in reality , however , the populated eigenstates undergo relaxation due to the interaction with the environment . to elucidate the effect of the relaxation on the quantum yield , we calculated the transition rates @xmath109 between eigenstates @xmath32 and @xmath110 using fermi s golden rule @xcite @xmath111 and @xmath112 where @xmath113^{-1}$ ] is the bose distribution at temperature @xmath114 k , @xmath115 is the boltzmann s constant , and @xmath116 is the spectral density of the bath modes representing low - frequency , non - reactive vibrational modes @xcite coupled to degree of freedom @xmath117 . following stock and co - workers @xcite
, we adopt an ohmic spectral density @xmath118 with @xmath119 , @xmath120 ev , @xmath121 , and @xmath122 ev .
the bath operators @xmath123 in eqs .
( [ relax1])-([relax2 ] ) are given by @xcite @xmath124 and @xmath125 .
figure 4(a ) shows the time dependence of the expectation values @xmath126 and @xmath127 in eq .
( [ populations ] ) obtained by propagating the rate equations parametrized by the transition rates given by eqs . ( [ relax1])-([relax2 ] ) with the initial condition @xmath128 , _
i.e. _ assuming fully incoherent excitation with natural light
. a substantial fraction of population at @xmath40 resides in the excited diabatic electronic states [ @xmath129 .
the interaction with low - frequency bath modes leads to dissipation of the electronic and vibrational energy , manifested in the decay of the excited - state populations @xmath130 and @xmath131 .
the decay is accompanied by a growth of the ground - state populations @xmath132 and @xmath133 shown by the solid lines in fig .
4(a ) . figure 4(b ) plots the time dependence of the quantum yield given by eq .
( [ y ] ) .
remarkably , the quantum yield shows only a weak time dependence , with deviations from the asymptotic value of 0.62 not exceeding 3% over the time interval studied ( 0 3 ps ) .
thus , while the individual _ cis _ and _ trans_-populations evolve in time , the value of the quantum yield , defined as their ratio via eq .
( [ qy1 ] ) , remains constant .
analysis of the system - bath coupling matrix elements [ eq . ( [ relax1 ] ) ] shows that the matrix elements involving the torsional degree of freedom @xmath55 are small compared to those of the coupling mode @xmath59 .
we can therefore expect that relaxation of the initial `` bright '' eigenstates populated by incoherent fc excitation ( see below and fig .
5 ) is driven by the interaction of the coupling mode with the bath oscillators . indeed , as shown in fig .
4(b ) , neglecting the @xmath55-component of the system - bath coupling ( dashed curve ) has little effect on the time variation of the quantum yield .
the dominant role played by the coupling mode in the relaxation process is at the heart of our arguments presented below that analyze the lack of time dependence of the quantum yield . in order to gain insight into relaxation dynamics , we plot in fig .
5 the transient linear absorption spectrum of the two - state , two - mode model [ e.g. the populations @xmath134 during the various stages of the relaxation process . at time zero
, an incoherent mixture of `` bright '' eigenstates is assumed to follow fc excitation from the ground state .
the three dominant `` bright '' eigenstates that account for over 40% of all @xmath40 excited - state population are @xmath135 , @xmath136 , and @xmath137 , and are focused upon below . as time proceeds ,
the eigenstates begin to relax due to the interaction with the phonon bath . at @xmath138
fs , the population is seen to be spread over three lower - lying manifolds of states , which are separated from the initial `` bright '' manifold by a constant energy gap .
the dominant eigenstates populated in the decay of the bright state @xmath135 , are shown in fig .
we observe that during the course of relaxation , the population `` branches out '' into different final eigenstates until it finally settles in a steady state characterized by a stationary eigenstate distribution shown in the lowermost panel of fig .
5 . the steady state defines the asymptotic ( @xmath139 ) limit of the quantum yield , and is thus of particular importance to the theoretical description .
according to fermi s golden rule ( [ relax1 ] ) , the relaxation rates are determined by ( a ) the magnitude of the system - bath coupling matrix element , and ( b ) the value of the spectral density at the transition frequency @xmath140 . the existence of multiple intermediate relaxation stages in fig .
6 is a consequence of a fixed cutoff frequency of the bath @xmath122 ev . because the ohmic spectral density function peaks at @xmath141 and decays quickly away from the maximum @xcite , relaxation to the eigenstates separated from the initial state by the energy window @xmath142 is allowed ( provided that the corresponding coupling matrix elements are non - zero ) , whereas relaxation to the eigenstates outside the energy window can not occur via single phonon - emission .
however , such lower - lying eigenstates are populated via multiple phonon emission ( or relaxation cascade ) , as illustrated in fig .
6 . as for the quantum yield ,
relaxation influences the quantum yield through the time dependence of the populations @xmath143 which changes the eigenstates that contribute to the evolving quantity : @xmath144 therefore , in order to understand the time dependence of the quantum yield , it is necessary to follow both the time evolution of the populations @xmath143 and their localization properties , as manifest in the matrix elements of @xmath145 and @xmath146 . to examine the localization
, we define the probability density @xmath147 for eigenstate @xmath32 projected onto the diabatic electronic state @xmath61 @xmath148 where @xmath149 are the diabatic components of the eigenstate @xmath32 described by the wavefunction @xmath150 where the direct - product basis @xmath151 is chosen in such a way as to diagonalize the torsional vibronic coupling hamiltonian ( [ model ] ) at @xmath152 ( no vibronic coupling ) . in this latter case
, the problem reduces to that of two non - interacting electronic states , and the eigenstates ( [ expansion1 ] ) factorize into products of basis functions that depend on the @xmath55 and @xmath59 coordinates @xmath153 this expression for the system eigenstates provides a reasonable zeroth - order approximation to the true eigenstates of the vibronic coupling hamiltonian ( [ model ] ) in the energy region below the conical intersection ( @xmath154 @xmath78 ) @xcite . here
, the low - lying eigenstates of the full vibronic hamiltonian are localized in their own potential wells @xcite ( as clearly observed in fig .
2 ) . however , as the energy increases above the conical intersection , the eigenstates become strongly mixed by the vibronic coupling , and delocalize over the whole configuration space .
an example of such behavior is shown in fig .
7 , which shows the probability density of the bright eigenstate @xmath135 together with that of the ground state @xmath155 and the eigenstates populated in the first few relaxation stages ( green arrows in fig .
the bright state has a large probability amplitude in the _
cis_-region of the first excited diabatic state , which maximizes its overlap with the ground _
cis_-state , to enhance excitation .
the lower - lying states @xmath156 and @xmath157 start to experience the repulsive wall of the @xmath158 diabatic potential in the _ cis_-region of configuration space ( see fig .
1 ) . as a result ,
the population is completely transferred away from the repulsive region to the ground _
cis _ and excited _ trans _ regions .
there is no apparent preference for either _ cis _ or _ trans _ regions during these initial relaxation stages due to the strong mixing of all degree of freedom : as shown in fig .
4(a ) , both @xmath21 and @xmath22 increase monotonically with time . after being transferred one step below below the initial state , the population continues to relax to lower - lying eigenstates while preserving its profile along the reaction coordinate .
this can be attributed to the particular form of the system - bath coupling , which primarily depends on the @xmath59 coordinate and hence does not strongly alter the @xmath55-profiles of the eigenstates coupled by the bath .
in particular , we verified that the amount of _ cis _ and _ trans _ character of most of the states involved in the relaxation process is approximately constant .
exceptions do occur because of the strong vibronic coupling , which may occasionally change the @xmath55 distribution of some eigenstates due to the coupling mediated by the @xmath59 coordinate .
however , this situation is an exception rather than the rule : as shown in fig .
8 , only two out of @xmath720 eigenstates involved in the first stage of relaxation ( @xmath159 and @xmath160 ) are dramatically different in their _ trans_-character ( as quantified by the matrix element @xmath161 ) from the other states . as a result ,
the quantum yield varies only weakly with time .
a dramatic change in the relaxation mechanism occurs when the eigenstate energy falls below that of the conical intersection .
figure 9 shows that the eigenstates to which relaxation then occurs become localized in their corresponding _
cis- _ and _ trans_- wells .
the population of the initial ( delocalized ) eigenstate @xmath162 relaxes to two eigenstates , one of which ( @xmath163 ) is strongly localized in the _ trans_-well and the other ( @xmath164 ) has more probability amplitude in the _ cis_-well ( see supplementary material @xcite ) . as the population gets partitioned between the two eigenstates , the quantum yield remains unaltered again due to the @xmath55-independent nature of the system - bath coupling , which tends to conserve the angular probability density of all eigenstates involved . finally , we consider the final stages of the relaxation process depicted in fig .
9 . these represent bath - induced transitions between localized eigenstates , and can be understood by noting that the eigenstates are given to zeroth order by eq .
( [ zeroth_order ] ) . for simplicity ,
let us assume that the value of the quantum yield before relaxation begins is determined by a single eigenstate @xmath165 .
after relaxation is over , the population is transferred to the eigenstate @xmath166 .
taking the matrix elements of @xmath146 and using eq .
( [ zeroth_order ] ) , we find @xmath167 because @xmath146 does not depend on @xmath59 , and similarly @xmath168 but the system - bath coupling operator does not depend on @xmath55 , and hence can not change the number of quanta in the torsional mode @xmath169 , so the right - hand sides of eqs .
( [ loc1 ] ) and ( [ loc2 ] ) are equal .
this result suggests that in the low - energy regime of localized eigenstates , the quantum yield of _ any _ process that depends on a single reaction coordinate should be time - independent , provided that the system - bath coupling does not depend on that coordinate .
this is clearly illustrated in fig .
9 : the localized state @xmath163 has 3 peaks along the @xmath59 coordinate , corresponding to 2 quanta in the @xmath59-mode ( @xmath170 ) .
the system - bath coupling changes @xmath171 from 2 to 1 and then from 1 to 0 , while leaving the @xmath55 distribution unchanged ( see also figs . 1 and 2 of supplementary material @xcite ) .
the final eigenstate @xmath172 is vibrationally `` cold '' with respect to the coupling mode , but remains highly excited along the torsional coordinate . a few closing remarks are in order concerning the asymptotic ( @xmath139 ) value of the quantum yield that corresponds to the population distribution shown in the lowermost panel in fig .
the steady state is obtained by propagating the rate equations of motion for the populations , which is a numerically efficient procedure that scales quadratically with the number of eigenstates @xcite .
the nature of the steady state is determined by the form of the system - bath coupling , and the initial conditions .
we emphasize that this steady - state solution does _ not _ have the form of the boltzmann distribution that would be normally expected of pauli - type rate equations in the limit @xmath139 @xcite .
rather , in accord with ref . , our steady - state distribution corresponds to a metastable state , which will eventually tunnel to the _ cis_-well ( at least if the well is one - dimensional ) . the large barrier height in the two - state two - mode model makes the tunneling timescale extremely long compared to any other timescale of interest in this system . as a possible direction of future research
, it would be interesting to explore whether such metastable steady states can be obtained without propagating the dynamical equations of motion .
if so , the determination of the quantum yield from the stationary eigenstates becomes a straightforward task via eq .
( [ qy1 ] ) with the matrix elements @xmath108 replaced by their steady - state values .
we have considered a time - independent approach to the quantum yield of _ cis - trans _ photoisomerization , here applied to model retinal in rhodopsin .
the need for this approach arises due to the recognition that natural processes take place in incoherent light ( e.g. , sunlight with a coherence time of 1.32 fs ) and environmental decoherence , which produce mixtures of stationary hamiltonian eigenstates .
here we have recast one of the standard time - dependent definitions of the quantum yield in terms of time - independent quantities , the eigenstates of the hamiltonian and associated dipole transition matrix elements . the quantum yield
is then shown to be a direct reflection of the _ cis _ vs. _ trans _ character of the individual stationary eigenstates of the system and the associated dipole transition matrix elements from the ground electronic state .
further , applied to a model of retinal , this approach gives excellent results for the quantum yield , fully in agreement with experiment .
interestingly , relaxation from the initially prepared stationary mixture does not alter the quantum yield , a consequence of both the _ cis / trans _ partitioning of the stationary states and the system - bath coupling in this well established minimalist model of retinal isomerization .
ideally , we should consider a full treatment of all modes of retinal in a proper rhodopsin environment , define the quantum yield and other observable properties , fit the retinal potential parameters to experiment , and compare time - dependent computational results to the stationary state results associated with incoherent light excitation .
however , such and extensive computational study is not required to support the main result of this paper , i.e. , that the stationary eigenstates of the system hamiltonian provide an alternative and important way to understand features affecting the quantum yield in incoherent light .
rather , we adopt the basic two - dimensional model of hahn and stock @xcite . restricting attention to this model
necessitates , by requirements of consistency , that if we adopt their two - dimensional potentials and associated system parameter fits to experimental data , that we also must maintain their definition of quantum yield , with the associated neglect of @xmath130 and @xmath131 .
alternative definitions of the quantum yield would have resulted in different values of the system parameter fits and different stationary eigenstates of the hamiltonian .
hence , the successful computational results demonstrated here do motivate a more extensive calculation of retinal hamiltonian eigenstates for a retinal model including all degrees of freedom@xcite .
such work is in progress .
* acknowledgments*. this work was supported by the natural sciences and engineering research council of canada , and by the us air force office of scientific research under contract numbers fa9550 - 10 - 1 - 0260 and fa9550 - 13 - 1 - 0005 .
table i. calculated quantum yields for retinal photoisomerization . both pre - averaged ( @xmath107 ) and post - averaged ( @xmath102 ) results are shown .
@xmath88 normalized lineshape function calculated within the 2d model ; @xmath173 normalized lineshape function based on the measured absorption spectrum of retinal in rhodopsin @xcite .
the time - dependent wavepacket result from ref . is given in the last column .
mathies , r. a. ; lugtenburg , j. in _ handbook of biological physics , volume 3 : molecular mechanisms in visual transduction , _ eds .
d. g. stavenga , w. j. degrip and e. n. pugh jr . , elsevier science press , pp .
55 - 90 ( 2000 ) .
polli , d. ; alto , p. ; weingart , o. ; spillane , k. m. ; manzoni , c. ; brida , d. ; tomasello , g. ; orlandi , g. ; kukura , p. ; mathies , r. a. ; garavelli , m. ; cerullo , g. conical intersection dynamics of the primary photoisomerization event in vision .
_ nature ( london ) _ * 2010 * , _ 467 _ , 440 - 443 .
kim , j. e. ; tauber , m. j. ; mathies , r. a. analysis of the mode - specific excited - state energy distribution and wavelenth - dependent photoreaction quantum yield in rhodopsin .
_ biophys .
j. _ * 2003 * , _ 84 _ , 2492 - 2501 .
zewail , a. h. ; orlowski , t. e. ; jones , k. e. radiationless relaxation in `` large '' molecules : experimental evidence for preparation of true molecular eigenstates and born - oppenheimer states by a coherent laser source .
usa _ * 1977 * , _ 74 _ , 1310 - 1314 .
mller , u. ; stock , g. flow of zero - point energy and exploration of phase space in classical simulations of quantum relaxation dynamics .
ii . application to non - adiabatic processes .
_ j. chem .
phys . _ * 1999 * , _ 111 _ , 77 - 88 . abe , m. ; ohtsuki , y. ; fujimura , y. ; domcke , w. optimal control of ultrafast _ cis - trans _ photoisomerization of retinal in rhodopsin via a conical intersection .
_ j. chem .
phys . _ * 2005 * , _ 123 _ , 144508 - 1 - 144508 - 10 .
leitner , d. m. ; kppel , h. ; cederbaum , l. s. statistical properties of molecular spectra and molecular dynamics : analysis of their correspondence in no@xmath174 and c@xmath174h@xmath175 .
_ j. chem .
phys . _ * 1996 * , _ 104 _ , 434 - 443 .
joubert - doriol , l ; ryabinkin , i. g. ; izmaylov , a. f. geometric phase effects in low - energy dynamics near conical intersections : a study of the multidimensional linear vibronic coupling model .
_ j. chem .
phys . _ * 2013 * , _ 139 _ , 234103 - 1234103 - 10 .
christopher , p. s. ; shapiro , m. ; brumer , p. overlapping resonances in the coherent control of radiationless transitions : internal conversion in pyrazine .
_ j. chem .
phys . _ * 2005 * , _ 123 _ , 064313 - 1 - 064313 - 9 .
doukas , a. g. ; junnarkar m. r. ; alfano r. r. ; callender r. h. ; kakitani t. ; honig b. fluorescence quantum yield of visual pigments : evidence of subpicosecond isomerization rates .
_ proc . natl .
_ * 1984 * , _ 81 _ , 4790 - 4794 . fig . 1 : ( upper panel ) ground and first excited diabatic potential profiles along the reaction coordinate @xmath55 for retinal photoisomerization .
the green upward arrow illustrates laser excitation , the downward arrows illustrate the partitioning of the eigenstates into _ cis _ and _ trans _
( by the projection operators , see text ) .
the _ cis _ and _ trans _ regions of configuration space are separated by the red dashed line .
( lower panel ) adiabatic pess for retinal as functions of the torsional coordinate @xmath55 and the coupling mode @xmath59 orthogonal to it .
the pess are obtained by diagonalizing the hamiltonian in eq .
( [ model ] ) ( without the kinetic energy term ) .
2 : frequency dependence of the stationary quantum yield . superimposed on the plot are the linear absorption spectrum of the 2d model ( green ) and the experimental absorption spectrum spectrum of retinal in rhodopsin adapted from fig . 2 of ref .
23 , both normalized to unity at their respective maxima .
the inset shows an expanded view of the frequency - dependent quantum yield in the region of maximum absorption at @xmath176 @xmath78 ( @xmath177 nm ) fig .
3 : wavelength dependence of the stationary quantum yield . diamonds
results for @xmath88 calculated from the two - state two - mode model ; dashed ( dash - dotted ) lines results obtained for @xmath107 ( @xmath102 ) and the experimentally measured @xmath74 @xcite , circles experiment .
the error bars are smaller than the size of the circles .
the inset shows the bare frequency - dependent quantum yield ( [ ybare ] ) calculated _ without _ the spectral lineshape function @xmath74 .
the experimental quantum yield stays constant below @xmath177 nm .
the results for @xmath88 are shown only in those spectral regions where @xmath88 does not vanish ( see fig .
2 ) . fig . 4 : ( a ) time dependence of _ cis _ and _ trans _ state populations ( see eq .
[ populations ] ) .
the excited - state populations @xmath23 and @xmath24 ( dashed lines ) decay in time due to the interaction with the environment ( see text ) .
( b ) time dependence of the quantum yield given by eq .
( [ y ] ) .
these results are obtained by solving the equations of motion for the diagonal elements of the density matrix parametrized by the transition rates given by eqs .
( [ relax1])-([relax2 ] ) .
the quantum yield remains constant ( within 3% of the asymptotic value of 0.62 ) over the time interval @xmath178 fs .
fig . 5 : snapshots of eigenstate populations @xmath143 . at @xmath40 ,
the bright eigenstates ( mostly 512 , 507 , and 508 ) are populated by fully incoherent , impulsive fc excitation ( see eq .
6 ) . at later times ,
interaction with the bath causes the population of the bright states to decay through several cascades ( middle panels ) .
the resulting steady - state eigenstate distribution is plotted in the lowermost panel .
fig . 6 : the network of dominant relaxation pathways starting from the dominant bright state @xmath135 ( the highest peak in the upper panel of fig . 5 ) .
green arrows show the most efficient pathway leading to the @xmath172 eigenstate ( the highest peak in the lower panel of fig . 6 ) .
see text for details .
fig . 7 : probability density ( eq . [ probden ] ) for the ground ( @xmath81 , upper panels ) and the first excited ( @xmath158 , lower panels ) diabatic electronic states as a function of the torsional coordinate @xmath55 and the coupling mode @xmath59 .
only the eigenstates above the conical intersection are shown .
note the similarity of the @xmath55-profiles of the eigenstates involved in the relaxation process , which implies no change in the quantum yield .
the lower left figure illustrates that _ cis / trans _ partitioning of delocalized eigenstates above the conical intersection ( marked by the star ) does not change qualitatively during relaxation .
note changes in color scale from panel to panel .
fig . 8 : localization properties of the eigenstates in the first relaxation cascade .
( upper panel ) : the populations of the first - cascade eigenstates at @xmath179 fs after the initial excitation .
( lower panel ) : the diagonal matrix element of the projection operator @xmath146 quantifying the amount of population in the _ trans_-region of configuration space .
note that all eigenstates significantly populated in the first stage of relaxation have similar localization properties with @xmath180 .
notable exceptions include states @xmath159 and @xmath160 with @xmath181 and @xmath182 .
same as in fig . 7
but for the eigenstates in the vicinity of and below the conical intersection . note the different localization properties of the eigenstates involved in the relaxation process .
the lower left figure illustrates the transition from delocalized to localized eigenstates below the conical intersection ( marked by the star ) .
note changes in color scale from panel to panel . | _ cis - trans _ isomerization in retinal , the first step in vision , is often computationally studied from a time dependent viewpoint .
motivation for such studies lies in coherent pulsed laser experiments that explore the isomerization dynamics .
however , such biological processes take place naturally in the presence of incoherent light , which excites a non - evolving mixture of stationary states . here
the isomerization problem is considered from the latter viewpoint and applied to a standard two - state , two - mode linear vibronic coupling model of retinal that explicitly includes a conical intersection between the ground and first excited electronic states .
the calculated quantum yield at 500 nm agrees well with both the previous time - dependent calculations of hahn and stock ( 0.63 ) and with experiment ( @xmath0 ) , as does its wavelength dependence .
significantly , the effects of environmental relaxation on the quantum yield in this well - established model are found to be negligible .
the results make clear the connection of the photoisomerization quantum yield to properties of stationary eigenstates , providing alternate insights into conditions for yield optimization . |
deuterium ( d or @xmath7h ) is the only element entirely produced by nuclear reactions in the first minutes after big bang ( wagoner et al .
the d yields are the most sensitive to the nuclear density at the nucleosynthesis epoch among the primordial light elements @xmath8he , @xmath9he and
@xmath10li , thus making the d abundance the most sensitive measurement of the baryon density in the universe ( wagoner 1973 , schramm & turner 1998 ) . deuterium is currently measured in the local interstellar medium ( ism ) , ( d / h)@xmath11 ( linsky et al . 1993 ) , but since whenever it is cycled through stars it is completely burned away , extrapolation to the primordial d / h value requires a modeling of the galactic chemical evolution .
direct d measurements of primordial material are thus of high interest . adams ( 1976 ) suggested that almost primordial d could be measured in low metallicity absorption line systems in the spectra of distant quasars ( qsos ) .
this was recently achieved for a few systems , but with conflicting results differing by almost an order of magnitude .
a few systems provide _ high _ d / h values ( see e.g. webb et al .
( 1997 ) who measure @xmath12 ) , while two other systems give a _ low _ abundance at @xmath13 ( burles & tytler 1998a,1998b ) .
an even lower d / h estimation was obtained by molaro et al .
( 1999 ) , further discussed by levshakov et al .
kirkman et al . ( 2000 ) measured an upper limit of @xmath14 .
the handful of d detections found so far does not allow a firm conclusion .
different arguments favour a low primordial d / h ratio : the possible contamination of the absorption lines and , on the modeling side , the results by tosi et al . ( 1998 ) which predict for a variety of chemical evolution scenarios and to be consistent with the galactic data a maximum decrease of the primordial d abundance by a factor of 3 .
the paucity of suitable absorption systems for accurate d / h measurements is due to the fact that only absorption line systems with simple velocity structures and with intermediate column densities allow the detection of the lines . at too low column densities
the lines are too weak for detection , whereas at high column densities the lines are normally washed out by the saturation of the line .
we show here for the first time that in the latter case the deuterium signature can be successfully detected through the higher members of the lyman series , when the target is a damped ly@xmath0 system ( @xmath15 ( ) @xmath16 ) at high redshift .
this approach was first suggested by khersonsky et al .
the spectra of qso 03473819 ( v=17.3 , @xmath17 ) were obtained during the commissioning of the ultraviolet - visual echelle spectrograph ( uves ) on the vlt 8.2 m kueyen telescope at paranal , in 1999 .
the instrument is described in dodorico et al .
two exposures of 4500 s each , covering the spectral range from 3650 to 4900 and from 6700 to 10000 were obtained with a resolution of @xmath18 km s@xmath19 and @xmath20 km s@xmath19 respectively
. the individual spectra were reduced using the uves data reduction pipeline implemented in the eso midas package .
the final spectrum reaches a s / n varying from 20 to 35 .
qso 03473819 shows a damped ly@xmath0 system at @xmath21 which has been studied in detail by centurion et al .
( 1998 ) , ledoux et al .
( 1998 ) and from hires - keck observations by prochaska & wolfe ( 1999 ) .
the uves observations however provide the first high - quality data in the uv ( @xmath22 ) and in the near - ir ( @xmath23 ) .
they allow the abundance measurements of new features such as o , p , ar and zn , in addition to the n , s , si and fe abundances measured in the previous studies .
full abundance analysis of the dla system at @xmath21 will be presented in a future paper . here
we focus on the deuterium detection and the d / h ratio measurement
. .column densities [ cols="<,^,<,^,^ " , ]
the absorption profiles of the dla system at @xmath21 are characterized by two dominating components ( 2 and 3 ) separated by about 20 km s@xmath19 with the red one slightly stronger than the blue as it can be seen from the non - saturated metal lines in fig .
[ low - ion ] .
the strong and saturated metal lines reveal that additional material in smaller amount is present redwards the two main features ( components 4 , 5 , 6 and 7 ) .
the region on the blue side of the main absorption components is sharp and relatively free from material with only one weak component ( component 1 ) at about @xmath24 km s@xmath19 . in total seven components
are needed to fit the metal lines absorption profiles , with two main components containing about 80% of the total column density per transition .
the dla system at @xmath21 is a very good candidate for the deuterium analysis since it shows a relatively simple velocity structure dominated by two strong components , a low metallicity of [ zn / h ] = @xmath25(x / h ) @xmath26(x / h)@xmath27 @xmath2 , indicating that the measured d / h will be representative of primordial d / h and the hydrogen lyman series lines can be followed down to ly12 .
we used a @xmath28 minimization routine fitlyman ( fontana & ballester 1995 ) in midas to fit voigt profiles to the observed absorption profiles , and obtain for each fitted absorption component the wavelength , the column density @xmath29 , the doppler parameter @xmath30 and the corresponding errors . in the case of high hydrogen column densities like in dla systems , we expect the neutral and low ionization metal lines to trace the h , therefore we model the lyman series ( ly@xmath31 to ly12 ) absorption profiles with three features corresponding to the two main components ( 2 and 3 ) and the reddest component ( 7 ) as determined from the metal lines ( fig .
[ low - ion ] ) .
the contribution of the weaker component 7 is required only to better constrain the fit on the red edge of the lyman lines .
the relative intensities of these three major hydrogen components were scaled with the metal lines assuming they have approximatively the same abundance ratios from component to component .
starting from this basis we obtained the final column densities and @xmath30-values ( free parameters ) by fitting simultaneously the lines of the lyman series and by assuming the same redshift for the three h components as for the metal lines .
the ly@xmath31 absorption profile provides a very good constraint to the total column density and the ly8 , ly9 , ly10 and ly12 profiles ( which are the ones free from strong contamination ) to the @xmath30-values . from the best fit ( fig .
[ lyb ] , fig .
[ ly ] , table [ hi - di ] ) we derived a total column density of @xmath15 ( ) @xmath32 in close agreement with the pettini et al .
( 1994 ) value of @xmath33 .
adding other components to the hydrogen fit with low column densities as the ones observed in the stronger metal lines , component 1 at @xmath24 km s@xmath19 and components 4 , 5 and 6 at slightly higher redshifts than the two main components , does not change significantly neither the total column density of the two main components nor the fit on the blue and red wings of the lyman lines . fig .
[ ly ] however clearly shows that the fit with three hydrogen components systematically fails to reproduce the edge of the blue wing in the higher members of the hydrogen lyman series , ly@xmath3 , ly8 , ly10 and ly12 , at about @xmath34 km s@xmath19 from the two h main components ( 2 and 3 ) which is the expected displacement of the corresponding d lines
. we then add to the model the deuterium by assuming its contribution only to the two main components ( 2 and 3 ) .
in the fitting procedure we assumed the same redshift for the two fitted components as for h and we left the column densities and the broadening parameters @xmath30 as free parameters .
the best fit performed over ly8 and ly10 improves the @xmath28 of the hydrogen fit by a factor of 3 .
it reproduces nicely the blue wing of the ly8 , ly10 and ly12 absorption profiles and is consistent with ly@xmath3 ( see fig .
[ ly - di-2 ] ) .
it gives a deuterium abundance of @xmath35 the error follows from the h and d fits and it includes the uncertainties in placing the continuum level and the errors on the @xmath30-values .
the column densities and the @xmath30-values of the two fitted components are given in table [ hi - di ] .
to check the stability of our result when the assumption that h and d mimic the metal structure is relaxed , we made a fit of the ly@xmath31 , ly@xmath3 , ly8 , ly10 and ly12 lines with three h and d components by keeping both their redshifts and column densities as free parameters .
we obtained a satisfactory fit with shifts of the components with respect to the values given in table [ hi - di ] of less than 4 km s@xmath19 .
the derived d / h ratio was comprised within our d / h error range .
a fit with only one strong h component and the corresponding d component gave also a d / h ratio within our d / h error range .
three arguments support the interpretation of the absorption feature detected on the blue wing of some of the higher members of the lyman series , ly@xmath3 , ly8 , ly10 and ly12 , as the absorption with the same two main components as detected in the metal lines of the dla system .
first , the absorption in the blue wing of the lyman series lines could also be explained by a hydrogen interloper associated with the damped ly@xmath0 system with a column density between @xmath36 and @xmath37 , a @xmath30-value between 15 and 25 km s@xmath19 and placed between @xmath38 and @xmath39 km s@xmath19 .
however , from the density distribution of the ly@xmath0 clouds in the forest at @xmath40 ( see e.g. kim et al .
2000 ) , the probability to have such a cloud at that position is smaller than 1/1000 .
the lack of any metal component at this velocity in the strong and saturated lines of the dla system ( see fig . [ low - ion ] ) provides an additional evidence for discarding this possibility .
secondly , the contamination on the blue wing of the lyman series lines by different interlopers at different redshifts which would mimic the same deuterium abundance as derived from the ly@xmath3 , ly8 , ly10 and ly12 lines is even more unlikely .
finally , the d and h column densities and broadening parameters @xmath30 resulting from the fits are consistent one relatively to other in the two fitted components ( 2 and 3 ) : the derived d / h ratios are the same and the @xmath30()/@xmath30 ( ) ratios are close to what is expected in the thermally dominated case .
this measurement of d is the first made in a dla system .
it shows that the dla systems with their low metallicity ism are a very promising class of absorbers for measurements of the d / h ratios at high redshifts , when it is possible to measure the higher members of the lyman series .
a systematic program of measurements using uves data is under way .
the derived d / h ratio of @xmath41 is close to the low values obtained by burles & tytler in lyman limit systems ( 1998a,1998b ) and it makes the claim of the primordial low d / h ratio more robust .
taken at face this ratio gives a baryon to photon ratio , @xmath42 , of @xmath43 ( burles et al .
this @xmath44 implies a helium abundance ( in mass ) of y@xmath45 = 0.2480 and a lithium abundance of @xmath46 which are both larger than presently allowed by observations of he in extragalactic regions and of li in halo stars ( izotov & thuan 1998 , bonifacio & molaro 1997 ) . on the other hand an @xmath44 of @xmath43 corresponds to a present - day baryon density of @xmath6 km s@xmath19 mpc@xmath19 ] which remains significantly lower than the @xmath48 derived from cmb anisotropy ( jaffe et al . 2000 ) .
we are indebted to the uves project team for the high quality of the spectra obtained early in the operation of the instrument .
we like to thank j.x .
prochaska for making available the profiles of the stronger metal lines and for comments on an earlier version of the manuscript . | we present the first d / h measurement in a damped ly@xmath0 ( hereafter dla ) system at @xmath1 towards qso 03473819 obtained from the uves - vlt spectra .
the dla absorber has a metallicity of [ zn / h ] @xmath2 and a relatively simple velocity structure , with two dominating components detected in several metal lines .
the hydrogen lyman series can be followed down to ly12 thanks to the high uv - blue efficiency of uves .
the best fit of the lyman series lines , and in particular of ly@xmath3 , ly8 , ly10 and ly12 , relatively free of local contamination , is obtained when the absorption is included in the two main components .
the measured deuterium column density yields @xmath4 close to other low d / h values from lyman limit systems .
the corresponding values for the baryon to photon ratio and the baryon density derived from d / h are @xmath5 and @xmath6 respectively . |
whereas the bulk behavior of homogeneous ( charged @xcite and uncharged @xcite ) colloidal suspensions is rather well understood , the situation for its inhomogeneous counterpart , such as that emerging in an adsorption process , is less clear .
potential applications of adsorption of charged colloidal particles ( macroions ) can vary from technological processes such as surface coating @xcite to biological material problems .
@xcite from a fundamental point of view , the tremendous long - ranged coulomb interaction that sets in represents a formidable theoretical challenge .
consequently , a deeper understanding of the phenomenon of macroion adsorption is justified and needed .
on one hand , experiments @xcite and the well known mean field gouy - chapman theory @xcite seem to nicely agree for the ion distribution of an aqueous monovalent electrolyte near planar charged interfaces , as long as non - specific forces as well as excluded volume effects are negligible .
on the other hand , if the solution contains highly multivalent and/or large sized ions , then the gouy - chapman theory may severely qualitatively fail .
a crucial missing ingredient in this theory is the inclusion of the finite size of the macroions that can lead to non - trivial phenomena , such as substrate s surface charge reversal , already with monovalent ions .
@xcite the lateral macroion - macroion electrostatic correlations that are also absent in a mean field theory can be attenuated ( and even become marginal ) at sufficiently high salt content , in contrast to excluded volume effects . in the past , gonzles - mozuelos and medina - noyola @xcite used integral equations to address the problem of macroions near repulsive / attractive charged walls interacting via an effective yukawa potential .
more recently , netz @xcite investigated thoroughly and analytically the behavior of macroions near charged interfaces , based on the debye - hckel ( dh ) theory , but ignoring its finite size .
thereby , the striking effect of surface charge amplification advocated here could not be captured by those approaches .
@xcite it is only very recently , that this phenomenon was reported by lozada - cassou and coworkers .
@xcite for a size - asymmetrical electrolyte , @xcite surface charge amplification was identified by solving numerically the modified gouy - chapman ( where cations and anions have different distances of closest approach to the interface ) .
intimately related to our work , the phenomenon of surface charge amplification in presence of macroions was also found by applying a sophisticated hypernetted chain / mean spherical approximation ( hnc / msa ) integral equation .
@xcite in this work , we present a simple model , where the finite size of the macroions is taken into account , to reveal the mechanisms governing macroion adsorption .
although our approach is less accurate than the one used by lozada - cassou and coworkers , @xcite it presents the nice advantage to be analytical and very intuitive .
the basic driving force of surface charge amplification is that ( spherical ) macroions tend to be surrounded by its counterions over _ its whole surface in a uniform manner_. this corresponds actually to the old classical thomson s sphere problem .
@xcite as long as the strength of the surface charge density of the oppositely charged substrate is low enough , a finite number of counterions of the macroions should stay in the vicinity of the interface ( see fig .
[ fig.model_setup ] ) , leading to a surface charge amplification .
our paper is organized as follows : our ( modified ) dh theory is explained in sec .
[ sec.dh-theory ] .
the results are presented in sec .
[ sec.results ] and followed by a brief summary ( see sec . [ sec.conclu ] ) .
our electrostatic model ( see fig . [ fig.model_setup ] ) resembles that employed by spitzer @xcite who studied monovalent size - asymmetrical ions near a wall . the negative substrate s surface charge density
is denoted by @xmath0 .
the macroions carry a positive central charge @xmath1 with @xmath2 representing the usual elementary charge and @xmath3 its valency .
excluded volume effects are taken into account via the distance of closest approach @xmath4 to the charged interface ( see fig . [ fig.model_setup ] ) .
our model corresponds somehow to the minimal correction for macroion size effect . having to deal with a _ finite _ concentration of macroions at contact with a reservoir , whose bulk value is given by @xmath5
, electroneutrality requires a bulk macroion s counterion concentration @xmath6 ( assuming monovalent point - like anions , i.e. @xmath7 ) .
those anions will be referred to as the _ coions _ of the substrate .
additional ( point - like ) monovalent counterions ( @xmath8 ) and coions are also considered with a bulk salt concentration @xmath9 . an intuitive and widely used way to connect self - consistently the electrostatic potential , @xmath10 , to the total charge density , @xmath11 ( with @xmath12 standing for the ionic species ) ,
is provided by the so - called poisson - boltzmann equation that reads : @xmath13,\end{aligned}\ ] ] where @xmath14 ( @xmath15 ) is the vacuum ( relative ) permittivity , @xmath16 is the reduced inverse temperature , and @xmath17 is the usual step ( or heaviside ) function .
this non - linear differential equation can only be solved numerically . however
a linearization of eq .
[ i.e. , dh approximation ] , valid for @xmath18 when @xmath19 and for @xmath20 when @xmath21 , permits an analytical treatment that is going to be discussed
. for the first diffuse region ( @xmath19 ) made up uniquely of the little monovalent ions ( see also fig .
[ fig.model_setup ] ) , the dh equation is given by @xmath22 with @xmath23 and @xmath24 , where we have introduced the bjerrum length @xmath25 and the dimensionless variable @xmath26 .
the corresponding solution reads : @xmath27 where @xmath28 . @xmath29 and @xmath30 are integration constants that are going to be determined after having applied the suitable boundary conditions . for the second diffuse region ( @xmath21 ) containing all the ions ( including the macroions - see also fig .
[ fig.model_setup ] ) the dh equation reads @xmath31 with @xmath32 $ ]
. the physically sound solution with vanishing electric field at @xmath33 corresponds to : @xmath34 where @xmath35 is a third integration constant .
our electrostatic model system is completely characterized once the integration constants @xmath29 , @xmath30 and @xmath35 appearing in eqs . and are specified . to do so , we apply the three following boundary and/or matching conditions : ( i ) the gauss law applied at the charged interface @xmath36 requires @xmath37 , where @xmath38 is the so - called gouy - chapman length .
( ii ) the zero intrinsic surface charge at @xmath39 imposes the continuity of the electric displacement @xmath40 and hence also that of the electric field @xmath41 [ i.e. , @xmath42 .
@xcite ( iii ) the continuity of the electrostatic potential at @xmath39 requires @xmath43 .
the resulting set of three equations can be readily solved and yields : @xmath44}_{\equiv q}. % \end{array } % \right .\end{aligned}\ ] ] note that in the limit of point - like macroions ( @xmath45 ) one recovers the well - known result @xmath46 . at this stage
all the relevant observables of the system can be in principle obtained within the framework of the dh theory .
a pertinent quantity that is appropriate to characterize the strength of the macroion adsorption is provided by the contact potential of interaction @xmath47 .
the latter corresponds to the external work accomplished upon bringing a macroion from infinity ( @xmath48 ) to contact ( @xmath39 ) : @xmath49 ( with @xmath50 denoting the @xmath51-component of the electric field ) . with the help of eq .
we get @xmath52 . using then eq . for @xmath35
, we obtain the following expression for @xmath47 : @xmath53.\end{aligned}\ ] ] as can be seen from eq .
, @xmath47 depends on many parameters , such as @xmath3 , @xmath54 , @xmath4 , @xmath55 and @xmath56 , @xcite making its understanding rather difficult . nonetheless , eqs . -
suggest that more insight on the adsorption behavior can be gained by considering two limits : ( i ) the high screening regime ( @xmath57 ) and ( ii ) the weak screening regime ( @xmath58 ) . without loss of generality
, we are going to explore these two limits . in the _ high screening regime _ ( @xmath59 ) , eq . can be approximated by : @xmath60 where the relation @xmath61 has been used .
interestingly , in this regime , @xmath47 depends only on the _ screening strength contrast _ @xmath62 .
one can also conveniently express @xmath47 as a function of the bulk concentrations @xmath5 and @xmath9 , which reads @xmath63 equation shows that @xmath47 is even independent of the bjerrum length @xmath64 , which means that the effective wall - macroion attraction is entropically ( or _ depletion _ ) driven . to better understand this phenomenon ,
we have sketched on fig .
[ fig.u_e_high_scr ] the profile of @xmath65 as well as profile of the reduced ( dimensionless ) electric field @xmath66 [ with @xmath67 for a set of typical parameters of charged colloidal suspensions : @xmath68 , @xmath69 ( corresponding to @xmath70 for an aqueous solvent ) , @xmath71 , @xmath72 , a ( fictive ) equivalent volume fraction @xmath73 set to @xmath74 , ( i.e. , @xmath75 with m standing for mole per liter units ) .
thereby we have @xmath76 .
the profile of @xmath77 reveals an _ unusual non - monotonic _ behavior near contact . this is a direct consequence of an accumulation of `` _ excess _ ''
coions in the macroion depleted zone ( @xmath78 ) leading to a _
weaker _ screening . in order to further characterize the adsorption behavior in the high screening regime
, we examine @xmath47 as predicted by eq . and
eq . for some typical experimental values of the parameters @xmath55 and @xmath56 of charged colloidal systems . in fig
[ fig.um_ns_high_scr ] , @xmath79 is plotted against the bulk salt concentration @xmath9 for different prescribed values of @xmath80 and @xmath81 .
only values of @xmath82 smaller than unity are shown such as to explore its behavior where the dh approximation is valid .
figure [ fig.um_ns_high_scr ] shows that beyond a certain threshold of salt content @xmath9 that is @xmath3-dependent , no adsorption occurs as signaled by a nearly zero value of @xmath47 . at relatively low enough salt concentration ,
adsorption is favored where @xmath47 increases with growing @xmath3 .
the latter point can be clearly understood by noticing that for typical values of the parameters of charged colloidal systems ( @xmath83 and @xmath84 ) we have @xmath85 . moreover , in the limit @xmath86 , the simple following relation @xmath87 holds .
we now address the _ weak screening regime _ characterized by @xmath88 .
in this situation eq . becomes : @xmath89 , \nonumber \\\end{aligned}\ ] ] where the dispersion relation @xmath90 has been used .
equation reveals that @xmath47 varies affinely with macroion size @xmath4 .
depending on the sign of @xmath91 , @xmath47 may either decrease or increase with @xmath4 .
for @xmath92 we obtain the following limit behavior : @xmath93 in this `` entropic '' limit @xmath94 increases linearly with @xmath4 , meaning that a finite wall - macroion attraction persists even at vanishing surface charge density @xmath0 due to excluded volume effect ( @xmath95 ) .
noticing that @xmath96 we deduce that the strength of the electric field at contact @xmath97 becomes larger than unity when @xmath98 is sufficiently large [ especially true for eq . ] .
this feature corresponds to an _ electric field ( or surface charge ) amplification _ at contact .
let us consider some typical experimental values of parameters representative of charged _ micellar _ systems ( see caption of fig .
[ fig.u_e_weak_scr ] ) .
thereby we have @xmath99 . to access the mechanisms of wall - macroion attraction in the weak screening regime ,
we have plotted the profiles of @xmath65 and @xmath77 in fig .
[ fig.u_e_weak_scr ] for various values of @xmath54 .
strikingly , in the macroion depleted zone , the electric field gets monotonically _ amplified _ ( from @xmath36 up to contact @xmath100 ) at ( very ) large @xmath54 [ corresponding to poorly charged interfaces ( @xmath101 ) - see fig .
[ fig.u_e_weak_scr ] ] .
this phenomenon is again due to an accumulation of `` excess '' coions in the macroion depleted zone , that leads here to a net surface charge that is more negative than @xmath0 .
it is important to remark , that surface charge amplification does not necessarily involve a subsequent charge reversal , as clearly indicated in fig .
[ fig.u_e_weak_scr ] .
in fact this charge - amplification phenomenon was already observed , although uncommented , in computer simulations [ see fig .
4(b ) in ref . @xcite and fig . 3 in ref .
to summarize , we have studied analytically within the framework of the debye - hckel theory the ( weak ) electrostatic adsorption of macroions at oppositely charged planar surfaces .
taking into account the crucial role of the finite size of the macroions , our model reveals non trivial adsorption driving forces . in the strong screening regime , the wall - macroion attraction strength at contact
is exclusively governed by the screening strength contrast @xmath62 . in the weak screening regime ,
the wall - macroion attraction strength can either decrease or increase with macroion size , depending on the surface charge density of the substrate .
in particular , at sufficiently small surface charge densities an effective electric filed ( or equivalently an effective surface charge ) amplification sets in .
all these adsorption mechanisms have a common feature , namely , the accumulation of excess coions in the macroion depleted zone .
the latter mechanism also explains the surface charge amplification recently reported by molecular dynamics simulations for dna - dendrimer complexation .
@xcite our findings could be experimentally verified by employing an ( extra ) ultra fine dispersion of charged particle tracers ( with a smaller size than the macroions and such as to nearly not modify the screening strengths @xmath55 and @xmath56 ) , whose density profile should reveal the electric filed one in the macroion depleted zone . | the adsorption of charged colloids ( macroions ) onto an oppositely charged planar substrate is investigated theoretically .
taking properly into account the finite size of the macroions , unusual behaviors are reported .
it is found that the role of the coions ( the little salt - ions carrying the same sign of charge as that of the substrate ) is crucial to understand the mechanisms involved in the process of macroion adsorption . in particular
, the coions can accumulate near the substrate s surface and lead to a counter - intuitive _ surface charge amplification_. |
bragg spectroscopy , using two laser beams , has proven to be a very successful tool for probing the structure factor of trapped atomic bec s @xcite . in this paper , we examine the analogous experiment on two - component fermi gases in both the normal and superfluid phases .
we calculate the bragg scattering rate for varying interaction strengths in both the weak coupling bcs and bec limits and in the strong coupling unitarity limit .
previously , we proposed measuring the inelastic ( stokes / anti - stokes ) scattering of an off - resonant laser beam on a two - component fermi gas @xcite . in the bcs limit ,
the intensity of the scattering in this stokes experiment exhibits a large maximum just below the critical temperature , @xmath0 , at which the gas becomes superfluid .
this effect is the light scattering analog of the hebel - slichter effect in conventional superconductors , a hallmark experimental test of bcs theory @xcite . here ,
we demonstrate that the same effect can be observed in a bragg scattering experiment .
we furthermore show that the pairing induced by the interactions can be readily detected in the bragg spectrum . in the strong coupling regime
, we discuss how the presence of a pseudogap in the normal phase has significant effects on the bragg spectrum .
the advantage of the bragg experiment as compared to the stokes experiment is that one varies in a controlled way the energy and momentum imparted to the atoms .
compared with recent radio frequency ( rf ) experiments which involve a third unoccupied hyperfine state @xcite , the bragg experiment involves only the two hyperfine states already present in the trap . the bragg experiment ,
avoiding complications due to non - trivial interaction effects between the third hyperfine state present in the rf experiment @xcite and the trapped gas , is thus simpler to interpret .
we consider a homogeneous gas of fermionic atoms occupying two hyperfine states , denoted by @xmath1 and @xmath2 .
we assume that the state @xmath2 has energy @xmath3 ( @xmath4 and @xmath5 in this paper ) above the state @xmath1 , and that the population of the two states is equal , @xmath6 .
the two atomic states interact with a coupling which can be characterized by a scattering length @xmath7 in a vacuum .
we ignore the interaction between atoms in the same hyperfine state since @xmath8-wave scattering is suppressed between two identical fermions .
bragg scattering is realized by a pair of laser beams with wave vectors @xmath9 and @xmath10 and frequencies @xmath11 and @xmath12 illuminating the gas .
an atom absorbing a photon from beam 1 and emitting a photon in beam 2 changes its energy - momentum by @xmath13 .
bragg scattering can be used to probe both the density and spin correlation functions .
pairing is predicted to have several effects on the frequency and momentum dependence of these correlation functions for @xmath14 @xcite . to probe the spin - flip correlation function ,
one tunes the two laser frequencies so that @xmath15 as illustrated in fig .
[ braggfig ] .
if the detuning , @xmath16 , from the optical transition to or from the upper level @xmath17 is large , population of the upper level can be neglected , and the effective light matter hamiltonian describing spin - flip bragg scattering can be written as @xmath18 , \label{braggterm}\ ] ] where @xmath19 , and @xmath20 describes the spatial profile of the laser beams and the dependence on the atomic states involved and their coupling to the electromagnetic field @xcite .
we neglect finite waist effects and take @xmath21 with @xmath22 . to measure the effects of pairing we focus on the spin - flip rate @xmath23 induced by the bragg lasers , where @xmath24 is here the number of particles in the level @xmath2 , and @xmath25 denotes a thermal average .
the heisenberg equation of motion for the operator @xmath24 is @xmath26= \int d^3 r\left[{\mathcal{i}}({\mathbf{r}},t ) \psi_\downarrow^\dagger({\mathbf{r}},t ) \psi_\uparrow({\mathbf{r}},t)-\rm{h.c.}\right].\ ] ] for sufficiently weak laser beams , @xmath27 can by calculated by linear response theory : @xmath28\rangle({\mathbf{q}},\omega ) } { \omega-\omega'}-c.c.\right)\nonumber\\ = -2{\mathcal{i}}^2{\rm im}\mathcal{d}({\mathbf{q}},\omega+i\eta ) , \label{braggterm}\end{gathered}\ ] ] with @xmath29 .
bragg scattering thus probes the fourier transform of the spin - flip correlation function @xmath30\rangle$ ] .
the response of the gas is usually detected experimentally by measuring the total momentum imparted on the gas @xcite ; the rate of momentum transfer is straightforwardly related to the scattering rate by @xmath31 where @xmath32 is the total momentum of the gas in the two hyperfine states .
in the weak coupling limit @xmath33 , the gas can be described by bcs theory .
the correlation function in ( [ braggterm ] ) is then a sum of contributions from normal and anomalous green s functions @xcite .
we obtain after some algebra , @xmath34 , \label{uvexpression}\end{gathered}\ ] ] with @xmath35 . here ,
@xmath36 and @xmath37 , etc . , with @xmath38 and @xmath39 as usual in bcs theory .
the quasiparticle energies are @xmath40 and @xmath41 with @xmath42 , where @xmath43 , @xmath44 is the chemical potential , and @xmath45 the bcs pairing gap .
the fermi functions are @xmath46 and @xmath47 , with @xmath48 .
the first term of ( [ uvexpression ] ) , describes scattering of quasiparticles while the last two terms describe the creation and annihilation of two quasiparticles
pair breaking processes . for an ideal gas in the normal phase , ( [ uvexpression ] ) yields in the limit @xmath49 , @xmath50 ( @xmath51 is the fermi temperature ) , and @xmath52 , @xmath53 where @xmath54 is the fermi energy , @xmath55 is the total density of the gas , and @xmath56 is the fermi momentum . the spin flip induced by bragg scattering can excite collective spin waves .
these modes are undamped in the normal phase for an attractive interaction @xcite . for a repulsive interaction in the normal phase ,
the spin waves are damped and they change the spectral shape of the particle - hole continuum
. the single particle approximation ( [ uvexpression ] ) does not include such collective modes . in order to include them ,
we sum bubbles to obtain the random phase ( rpa ) approximation @xcite , @xmath57 with @xmath58 the scattering matrix for @xmath59 .
note the @xmath60 sign in the denominator due to the spin flip .
we first consider the temperature dependence of the scattering rate .
an attractive feature of the stokes experiment is that one can observe a strikingly large maximum in the scattered signal just below the critical temperature @xmath0 @xcite .
we now demonstrate that the same effect is present for the bragg scattering experiment in the experimentally realizable parameter range , @xmath61 , @xmath62 , and @xmath63 .
figure [ tdepfig ] shows the scattering rate ( [ braggterm ] ) as a function of @xmath64 in the weak coupling limit , in units of the normal phase rate , which is constant for @xmath50 . for illustration ,
we have taken @xmath65 and @xmath66 . solving the bcs equations as a function of @xmath64
, we obtain the scattering rate from eqs .
( [ uvexpression ] ) and ( [ rpa ] ) .
the critical temperature is @xmath67 .
we plot the results for frequency differences between the two laser beams @xmath68 , @xmath69 , and @xmath70 .
the numerical normal phase results agree with ( [ nexact ] ) , confirming the numerical accuracy of the calculations . as fig .
[ tdepfig ] shows , the scattering rate in the superfluid state has a large maximum below @xmath0 for @xmath71 .
this maximum is due to a large quasiparticle scattering rate arising from the increased density of states at the fermi energy in the superfluid phase just below @xmath0 , @xmath72 . for @xmath71 , both scattering states @xmath73 and @xmath74 in eq .
( [ uvexpression ] ) can be in the region around the fermi surface of a large density of states yielding an increased scattering rate .
the maximum in the scattering rate depends crucially on the coherence factors for quasiparticle scattering in ( [ uvexpression ] ) adding , viz . , @xmath75 , for a spin flip .
for bragg scattering into the same hyperfine state , the coherence factor is @xmath76 at the fermi surface and the increased density of states would not lead to a maximum in the scattering rate . for @xmath77 ,
the scattering rate is suppressed by the increasing gap @xmath78 since the density of quasiparticles available for scattering scales as @xmath79 . for @xmath80
the lasers can not break cooper pairs .
the physical reason for the coherence factors adding ( @xmath75 ) for spin - flip scattering is that the quasiparticles carry the same spin in the superfluid as in the normal phase . on the other hand , in scattering without spin flip , which couples to the particle density
, the particle number carried by the quasiparticles is suppressed at the fermi surface , since a quasiparticle is a superposition of a hole and a particle . in this case
the coherence factors subtract ( @xmath81 )
. for increasing @xmath82 , the peak in the rate below @xmath0 decreases and eventually disappears for @xmath83 . from fig .
[ tdepfig ] , we see that the scattering rate in the superfluid state is always smaller than in the normal state when @xmath84 ( green dashed and black dash - dotted lines ) .
this is because both scattering states can not be in the region around the fermi surface where the single particle density of states is increased .
bragg scattering does not probe the increased density of states for the superfluid state for higher frequency as effectively as for @xmath85 , and the resulting rate is decreased compared to the normal state .
note that the kink in scattering rate just below @xmath0 for @xmath86 , at which point @xmath87 , is physical .
below this temperature , the gap is too large to allow the creation of two quasiparticles and the second term in ( [ uvexpression ] ) vanishes .
finally , fig .
[ tdepfig ] , for @xmath88 illustrates the case of large energy transfer , @xmath89 .
contrary to the case @xmath90 , the rate is finite for @xmath14 as the energy transfer is large enought to break pairs and the second term in ( [ uvexpression ] ) is finite . for @xmath91 , the scattering rate approaches that of the normal state , as expected .
we conclude that there is a peak in the scattering rate for @xmath92 and @xmath71 .
also , the peak disappears unless @xmath93 and @xmath94 ; otherwise , both scattering states can not be in the region of increased density of states .
we now consider the frequency dependence of the bragg scattering rate .
figure [ wdepfig ] shows the rate from ( [ rpa ] ) as a function of @xmath82 for the gas in the weak coupling limit with @xmath65 and @xmath95 .
the scattering rate is depicted for various temperatures below and just above @xmath0 where the gas is normal .
for comparison , we also plot the result of the single particle approximation ( [ uvexpression ] ) in the normal phase .
the large peak in the normal phase response just above the single particle continuum arises from the spin wave present above the particle - hole continuum for an attractive interaction .
the finite width of the peak is an artifact of our having included a small imaginary part in the frequency @xmath82 for computational reasons .
we see from fig .
[ wdepfig ] that when the gas becomes superfluid , the spin wave vanishes .
this is because only potential flow , i.e. , density fluctuation modes , is possible in the superfluid @xcite .
other modes such as spin waves require a normal component to be present . in a trapped system
this effect leads to the damping of the spin - dipole mode in the superfluid phase @xcite . in the superfluid phase ,
the particle - hole spectrum disappears and is replaced by a low frequency gap , which reflects the pair breaking energy described by the second term in ( [ uvexpression ] ) . at zero temperature
the gas does not respont for @xmath80 , the minimum energy required to break a pair .
the pairing gap can thus be detected directly as a gap of @xmath96 in the bragg spectrum . by contrast , in the rf experiment of @xcite the shift in the spectrum scales as @xmath97 , since one transfers particles to a third hyperfine state which in general interacts differently with the two paired hyperfine states @xcite .
we now consider bragg scattering in the weak coupling bec regime @xmath98 . in this limit , the molecules are tightly bound with a radius much smaller than the average interparticle distance , and are weakly interacting both with each other and with unbound atoms @xcite .
the system can therefore to a good approximation be regarded as an ideal gas of point molecules and atoms in thermal equilibrium .
the molecules bose condense below the critical temperature @xmath99 . the bragg scattering induced by ( [ braggterm ] )
then either breaks a molecule by flipping the spin of a bound @xmath100 atom , thereby creating two free @xmath101 atoms , or it simply flips the spin of a free @xmath100 atom .
the fermi golden rule result for bragg scattering rate into the free atomic @xmath101 states for such a system is then a sum of three terms , @xmath102 the scattering from the unbound atoms is given by @xmath103 where @xmath104 , @xmath105 , @xmath106 , and @xmath107 .
the rate from a bec of molecules with density @xmath108 is @xmath109 with @xmath110 the energy of a molecule with zero momentum relative to the bottom of the free particle continuum of a pair of @xmath100 and @xmath101 atoms , and @xmath111 the matrix element of ( [ braggterm ] ) ( excluding @xmath112 ) between a molecule at rest and a pair of free @xmath101 atoms with momenta @xmath113 and @xmath114 . the scattering rate from thermal molecules
is @xmath1152\pi\delta(\omega+e_b+\frac{p^2}{4m}-\epsilon-\epsilon ' ) , \label{bragtherm}\end{gathered}\ ] ] with @xmath116 the bose distribution , and @xmath117 the energy of molecules with momentum @xmath118 .
the molecule breaks into two @xmath101 atoms with momenta @xmath119 and @xmath120 and energies @xmath121 and @xmath122 .
the matrix element in ( [ bragtherm ] ) is , from galilean invariance , independent of @xmath123 . to proceed ,
we model the molecule - free atom pair matrix element @xmath124 as follows .
we write state of a molecule at rest as @xmath125 , with @xmath126 the vacuum and @xmath127 creating an atom with momentum @xmath128 in hyperfine state @xmath129 .
the fourier transform of the molecular wave function is @xmath130 .
the spin - flip process illustrated in fig.[braggfig ] breaking the molecule and adding momentum @xmath131 creating two free @xmath101 atoms with momenta @xmath113 and @xmath114 corresponds to @xmath132 .
however , the process @xmath133 connects to the same final state , and the two matrix element must be added coherently in the golden rule expression .
we thus find the matrix element @xmath134 ( where the factor @xmath135 avoids double counting ) .
the bcs wave function correctly describes a condensate of non - overlapping bosons ( cooper pairs ) with binding energy @xmath110 in the bec limit @xmath136 . to connect this wave function to the above analysis
, we note that in the bcs expression the bragg scattering ( [ uvexpression ] ) , it is the second term that describes the breaking of cooper pairs .
thus , bcs theory yields @xmath137 . in the bec limit
, we have @xmath138 and @xmath139 with @xmath140 .
bcs theory therefore yields @xmath141 for the fourier transform of the molecular wave function and @xmath142 in the bec limit .
this is the asymptotic bound state wave function for a potential with scattering length @xmath7 as expected .
we therefore use the bec limit of the bcs result for the matrix element @xmath143 .
note that we neglect any closed channel components deep in the bec regime .
the populations of the molecular and atomic states are found by solving the atom - molecule equilibrium problem ; the density of the thermal molecules , the molecular bec and the thermal atoms are shown in fig .
[ densfig ] for varying @xmath64 . in fig .
[ molbraggfig ] , we plot the scattering rate as a function of @xmath82 obtained from ( [ molbragg ] ) for a molecule energy @xmath144 ( corresponding to @xmath145 ) and for momentum transfer @xmath66 .
the scattering rate is shown in fig .
[ molbraggfig ] for a pure molecular bec at @xmath14 ( a ) and for thermal molecules at @xmath146 above the bec critical temperature @xmath99 . for @xmath14 ,
the rate is given by ( [ bragbec ] ) and the threshold for the scattering is given by @xmath147 the minimum energy required to break a @xmath148 molecule and create two @xmath101 atoms with momenta @xmath149 and @xmath150 .
the rate from the thermal molecules at @xmath146 obtained from ( [ bragtherm ] ) does not differ significantly from that of the molecular bec at @xmath14 .
this is because both temperatures are so low that there are essentially no atoms present , as one sees in fig .
[ densfig ] .
the only effect of finite temperature is then a small shift @xmath151 in the threshold for scattering , from ( [ bragtherm ] ) .
thus , the onset of bose - einstein condensation of molecules has only minor effects on the bragg scattering rate since @xmath152 in the bec limit . in fig .
[ molbraggfig ] ( b ) we plot the scattering rate for @xmath153 , at which most of the atoms are still bound as molecules but some free atoms are present , as can be seen from fig . [ densfig ] .
the bragg scattering consists of a signal down to @xmath154 from the free atoms , given by ( [ atombragg ] ) , and a signal from the thermal molecules for @xmath155 , given by ( [ bragtherm ] ) . note that even though the atomic signal at small @xmath156 is much larger than the molecular signal at high @xmath82 , the contribution to the f - sum rule from the molecules is much larger than that from the free atoms .
for higher temperatures , the molecular signal of course disappears with the vanishing molecule density , and the bragg scattering is given by that of the free atoms only .
the scattering length diverges , @xmath157 , at the feshbach resonance . in this unitarity limit
, the interactions are strong and considerably modify the properties of the gas .
an important characteristic of the fermi gas at unitarity is the presence of a pseudogap in the normal phase close to @xmath0 ; i.e. , the suppression of the single particle dos around the fermi surface @xcite .
a pseudogap has been observed experimentally for high @xmath0 superconductors where it is presently subject to intense investigations @xcite . since a gap in the spectrum of single particle excitations naturally leads to a gap in the low temperature bragg spectrum ( figs .
[ wdepfig]-[molbraggfig ] ) , we now examine whether the pseudogap can be detected in bragg spectroscopy .
we calculate the bragg spectrum at unitarity using a minimal many body theory that includes the correct two - particle physics leading to the resonance . for the single channel problem
, such a theory including pair fluctuations in the ladder approximation was originally developed many years ago @xcite and later examined in great detail by several authors @xcite .
this approach was then extended to the multichannel case relevant for atoms interacting via a feshbach resonance , treating the bound state as a point boson @xcite , and taking into account that the bound state is a composite two - fermion object @xcite .
the atoms are described by the atomic propagator @xmath158 with @xmath159 .
the atom self - energy is given in the ladder approximation by @xmath160\label{selfenergy}\ ] ] where @xmath161 , @xmath162 , the trace denotes a sum over matsubara frequencies @xmath163 and integration over @xmath128 , and @xmath164 is the many - body scattering matrix depending on the center of mass @xmath165 and relative momenta @xmath166 of the two scattering particles @xcite . in the limit of a broad resonance ,
the effective interaction mediated by the feshbach molecule can be regarded as instantaneous with a resonant scattering length @xmath167 at unitarity @xcite . neglecting the frequency and momentum dependence of the effective interaction coming from the molecular state
, we recover the single channel theory scattering matrix @xmath168 the pair propagator is , as usual , @xmath169 here , @xmath170 , @xmath171 , @xmath106 , and @xmath107 . the structure of eqs .
( [ selfenergy])-([pairprop ] ) is indicated in fig .
[ feynfig ] ( a ) . to obtain the bragg spectrum
, we need to evaluate the atomic propagator at frequency @xmath172 .
( from here on we do not write the infinitesimal positive imaginary part of the frequency explicitly . ) from ( [ selfenergy ] ) , we obtain @xmath173 .
\label{rselfenergy}\end{gathered}\ ] ] here @xmath174 , @xmath175 , and @xmath176 and @xmath177 are the spectral functions for non - interacting atoms and the pair propagator , respectively .
we evaluate the real and imaginary parts of @xmath178 directly from ( [ rselfenergy ] ) .
we also calculate the real part from the imaginary part using a kramers - kronig relation , and find that the two calculations agree to a high accuracy , confirming the consistency of the numerical calculations . in order to evaluate medium effects
, we need to determine @xmath0 and the chemical potential , @xmath44 .
we determine @xmath44 by calculating the thermodynamic potential @xmath179 at unitarity in the ladder approximation .
the correction @xmath180 to the ideal gas result can be expressed as an integral over the phase of the on - shell scattering matrix @xmath164 @xcite : @xmath181 with @xmath182 , @xmath183 defined below ( [ pairprop ] ) .
the angle @xmath184 is defined by writing @xmath185 . equation ( [ omega ] ) reduces to the virial expansion for @xmath179 at high temperatures .
the chemical potential @xmath44 is then determined by @xmath186 . in order to calculate the spin - flip correlation function consistently
, one needs in principle to choose a functional @xmath187 $ ] , generate the self - energy as @xmath188 , and then generate the correlation functions from the kernel @xmath189 @xcite .
the diagrams relevant for the ladder approximation are given in fig .
[ feynfig ] ( b ) where all propagators are full green s functions , @xmath190 , including those in @xmath164 .
the conserving spin - flip correlation function thus obeys an integral equation with @xmath164 and the full @xmath190 s .
the imaginary part of the first term in this equation is @xmath191 with @xmath192 , @xmath193 , and @xmath194 and @xmath183 defined as in ( [ rselfenergy ] ) .
the spectral function @xmath195 describes the excitations of the interacting atoms .
we calculate the bragg scattering rate at the critical temperature @xmath0 where we expect the pseudogap effects to be most pronounced . from the thermodynamic potential ( [ omega ] ) , we obtain @xmath196 and can calculate the critical temperature from the thouless criterion @xmath197 .
we then evaluate the atom self - energy from ( [ rselfenergy ] ) and from this obtain the spectral function . to understand the bragg scattering rate at unitarity
, we first analyze the effects of the interactions on the single particle spectrum , described by the spectral function , @xmath198 . in fig .
[ spectralfig ] , we plot @xmath198 at @xmath0 for various momenta , measured in units of @xmath199 .
we have chosen parameters corresponding to a resonant interaction with @xmath200 and a negligible effective range .
for this set of parameters , we find @xmath201 with @xmath202 in good agreement with other bec - bcs crossover results based on a similar approximations @xcite .
the spectral functions obey the sum rule @xmath203 to a very good approximation , providing another important check of the numerical calculations . in fig .
[ spectralfig ] , we clearly see the double peak structure of @xmath198 with the spectral weight at @xmath154 significantly suppressed for all @xmath204 . this suppression is the characteristic feature of the pseudogap .
we refer to the peak at @xmath205 as the _ hole peak _ at frequency @xmath206 , and the peak at @xmath207 as the _ particle peak _ at @xmath208 .
the positions of the two peaks found by fitting to a double lorentzian ( normalized such that its total spectral weight is @xmath209 ) are plotted in fig .
[ posfig ] ( a ) .
we see that the distance of the maxima to @xmath154 reaches a minimum : @xmath210 at @xmath211 , providing a qualitative value for the pseudogap .
the spectral weight of the hole peak decreases and that of the particle peak increases with increasing @xmath204 .
this behavior is illustrated in fig .
[ posfig ] ( b ) where we plot the spectral weights @xmath212 of the particle peak and the hole peak @xmath213 obtained by fitting to the double lorentzian . for large @xmath214 ,
the spectral function approaches that of an ideal gas with a single peak at @xmath215 as expected .
the behavior of the spectral function at unitarity as described here is in general agreement with the analysis of ref .
the spectral function behaves qualitatively as the bcs spectral function @xmath216 with two peaks located around @xmath154 .
there are important differences from bcs theory though .
the two peaks for the spectral function in the unitarity limit at @xmath0 have large widths ( which are unequal for the particle and hole peaks ) and the spectral density is suppressed but finite at @xmath154 .
the pseudogap is a _ suppression _ of spectral density and not a real gap . also , the positions of the two peaks are not symmetric around @xmath154 and the minimum gap is not at @xmath217 but slightly above , as can be seen from fig .
[ posfig ] .
we now ask whether the pseudogap behavior of the spectral function at unitarity has consequences for the bragg spectrum . in fig.[braggstrong ] , we plot @xmath218 for @xmath219 and @xmath220 . the @xmath221 s are obtained from ( [ imd0uni ] ) using the spectral function calculated from ( [ rselfenergy ] ) .
we evaluate the ( computationally demanding ) multidimensional integrations using a monte carlo routine .
the solid line in fig .
[ braggstrong ] is obtained by using the double lorentzian fit @xmath222 for the spectral function in ( [ imd0uni ] ) . here , the position of the hole @xmath223 and particle @xmath224 peaks , the width and the weight @xmath225 are calculated as functions of momentum @xmath204 using approximate functions obtained from fitting @xmath198 to the double lorentzian for various @xmath204 .
the approximate functions for @xmath223 , @xmath224 , and @xmath225 are given as dashed lines in fig .
[ posfig ] .
once the approximate double lorentzian form @xmath222 is obtained , the integration in ( [ imd0uni ] ) is much faster than when using the spectral function @xmath198 obtained from the full ladder calculation .
we see from fig .
[ braggstrong ] that the double lorentzian model @xmath222 yields results for @xmath226 in good agreement with the full numerical calculation based on @xmath198 . from this
, we conclude that the pseudogap spectrum close to @xmath0 can be well described by a simple double peak spectral function . comparing to the ideal gas scattering rate for the same density and temperature ,
also plotted in fig .
[ braggstrong ] , we see that interactions suppress the spectral weight of @xmath218 at low @xmath82 and push it to higher frequencies .
this is a direct effect of the pseudogap .
as opposed to the real gap in the @xmath14 bragg spectrum , the spectral density is however finite at small @xmath82 .
this is as expected since the single particle spectral density is suppressed but finite for @xmath154 ( fig .
[ spectralfig ] ) .
also , @xmath227 is comparable to the pseudogap @xmath228 so thermal excitations are present . from ( [ imd0uni ] )
, we see that that both effects make @xmath226 finite for small @xmath82 .
the approximation @xmath229 to the spin - flip correlation function is not conserving and does therefore not obey the f - sum rule . to obtain a conserving approximation
, we need to solve the integral equation in fig .
[ feynfig ] ( b ) .
this will change the quantitative form for @xmath230 .
the solution is however an iteration with the @xmath164 matrix of @xmath229 .
we therefore expect @xmath231 to have the same qualitative characteristics as @xmath218 , i.e. , a suppressed but finite spectral weight at low @xmath82 .
solving the fully self - consistent equation for the spin - flip correlation function is beyond the scope of this paper @xcite . in the bcs limit
, we see that the emergence of a gap in the single particle spectrum leads to a maximum in the bragg scattering rate just below @xmath0 ( fig . [ tdepfig ] ) .
one could therefore expect a similar maximum to appear at unitarity at a temperature @xmath232 defined as the temperature where the pseudogap emerges .
however , the presence of a maximum requires the spectral function to have well - defined sharp quasiparticle peaks like in the bcs case ( [ abcs ] ) .
the pseudogap spectral function at unitarity ( fig .
[ spectralfig ] ) on the other hand has peaks of width @xmath233 .
there will therefore be no maximum in the bragg scattering rate for @xmath234 due to the emergence of the pseudogap @xcite .
we conclude that the pseudogap present in the normal phase close to @xmath0 at unitarity leads to a significant suppression of the low frequency bragg scattering rate .
the rate remains however finite since the pseudogap is only a suppression of the single particle spectral density , and because @xmath0 is comparable to the value of the pseudogap .
the emergence of a pseudogap with lowering @xmath64 does not give rise to a maximum in the scattering rate , as opposed to the bcs case .
this is because the quasiparticles are strongly damped at unitarity .
we briefly comment on the possible experimental observation of the effects described in this paper .
a typical fermi energy for trapped atomic gases is @xmath235hz .
thus , we require a frequency resolution of the bragg experiment @xmath236hz .
likewise , a typical fermi momentum of the gas is @xmath237m@xmath238 which is the same order of magnitude as the wave numbers for optical lasers .
this means that the resulting angle between two laser beams needed to obtain a momentum transfer of @xmath239 is reasonable .
the results presented in this paper are for a homogenous system . from the thomas - fermi approximation
, we expect the presence of the trapping potential to smooth out features like the gap in the bragg spectrum . to avoid such smoothening , one could focus the laser beams on a smaller section of the cloud where the gas can be regarded as homogeneous . in this case , the results presented in the present paper become directly applicable .
we are grateful to cindy regal and chris pethick for helpful discussions .
we thank the aspen center for physics where parts of this research were carried out .
this research was also supported in part by nsf grants phy03 - 55014 and phy05 - 00914 .
d. m. stamper - kurn , a. p. chikkatur , a.grlitz , s. inouye , s. gupta , d. e. pritchard , and w. ketterle , phys.rev .
lett . * 83 * , 2876 ( 1999 ) ; j. steinhauer , r. ozeri , n. katz , and n. davidson , phys .
lett . * 88 * , 120407 ( 2002 ) .
a possible approach is via quantum monte carlo .
see , e.g. , calculations of the spin susceptibility in the pseudogap regime of the bcs - bec crossover by m. randeria , n. trivedi , a. moreo , and r. scalettar , phys .
lett . * 69 * , 2001 ( 1992 ) , and n. trivedi and m. randeria , phys .
. lett . * 75 * , 312 ( 1995 ) . | we propose a bragg spectroscopy experiment to measure the onset of superfluid pairing in ultracold trapped fermi gases . in particular , we study two component fermi gases in the weak coupling bcs and bec limits as well as in the strong coupling unitarity limit .
the low temperature bragg spectrum exhibits a gap directly related to the pair - breaking energy .
furthermore , the bragg spectrum has a large maximum just below the critical temperature when the gas is superfluid in the bcs limit . in the unitarity regime , we show how the pseudogap in the normal phase leads to a significant suppression of the low frequency bragg spectrum . |
the determination of the longitudinal polarization of the electron beam is one of the dominant systematic uncertainties in any parity violating electron scattering ( pves ) experiment . in order to achieve the desired high precision , the polarization of the electron beam
must be monitored continuously with an uncertainty of @xmath00.5% .
these ambitious goals can be achieved if multiple independent and high precision polarimeters are used simultaneously .
in addition to being precise , the polarimeters must be non - invasive and must achieve the desired statistical precision in the shortest time possible .
compton and mller polarimeters are typically the polarimeters of choice for these experiments and are essential to achieve the desired precision .
however , a complimentary polarimetry technique based on the spin dependence of synchrotron radiation , referred to as `` spin - light , '' can be used as a relative polarimeter .
a spin - light polarimeter could provide additional means for improving the systematic uncertainties and when calibrated against a compton / mller polarimeter it could provide a stable continuous monitoring of the beam polarization .
we develop the conceptual design for a continuous polarimeter based on `` spin - light '' .
the proposed spin - light polarimeter can achieve statistical precision of @xmath0 1% in measurement cycles of less than 10 minutes for 4 - 20 gev electron beams with a beam currents of @xmath1 100 @xmath2a .
.a comparison of the compton , mller and spin - light polarimeters .
[ cols="<,<,<",options="header " , ]
spin light based polarimetry was demonstrated over 30 years ago , but has been ignored since then .
a spin - light polarimeter has several advantages over conventional polarimeters and when used in conjunction with a compton polarimeter it could help provide a new benchmark for precision polarimetry
. the 11 gev beam at jlab or the electron beam at a future eic would be well suited for spin light polarimetry and such a polarimeter would help achieve the @xmath0 0.5 % polarimetry desired by experiments approved for the 12 gev era and proposed for the eic . a 3 pole wiggler with a field strength of 4 t and a pole length of 10 cm would be adequate for such a polarimeter .
a dual position sensitive ionization chambers with split anode plates is ideally suited as the x - ray detector for such a polarimeter .
the differential detector design would help reduce systematic uncertainties .
this work was supported in part by the u.s .
department of energy under contract # de - fg02 - 07er41528 , by the eic detector r&d grant from brookhaven national lab .
one of us ( p.m. ) would also like the thank the jefferson science associates for a jsa fellowship . 1 m. hauger
_ et al . _ ,
a462 * , 382 ( 2001 ) . to be reported for the polarization measured during the qweak experiment at jlab , hall - c .
p. c. rowson , d. su , and s. willocq , ann .
nucl . part .
sci . * 51 * , 345 ( 2001 ) ; m. woods , slac - pub-7319 ( 1996 ) .
j. allison _
et al . _ ,
inst . and meth .
* a506 * , 250 ( 2003 ) ; j. allison _
et al . _ , ieee trans . in nucl .
science * 53 * , 270 ( 2006 ) .
i. p. karabekov , r. rossmanith , proc . of the 1993 pac ,
washington , v. 1 , p. 457
( 1993 ) ; i. p. karabekov and s. i. karabekian , proceedings of 5th european particle accelerator conference ( epac 96 ) , sitges , spain , 10 - 14 jun 1996 , pp 1743 - 1745 ( 1996 ) ; a. v. airapetian , r. o. avakian , i. p. karabekov , e. l. saldin , and m. v. yurkov , proc . of the spin-96 , amsterdam , the netherlands , vol.1 , p762 ( 1996 ) .
d. d. ivanenko , i. pomeranchuk , ya zh
* 16 * , 370 ( 1946 ) ; j. schwinger , phys . rev . * 75 * , 1912 ( 1947 ) .
g. a. schott , ann . phys . * 24 * , 635 ( 1907 ) ; a. a. sokolov and i. m. ternov , zh .
. fiz . * 31 * , 373 ( 1956 ) , sov .
jetp * 4 * , 396 ( 1957 ) .
f. r. elder , r. v. langmuir and h. c. pollock , phys . rev . * 74 * , 52 ( 1948 ) .
m. yu ado and p. a. cherenkov , sov .
. dokl . * 1 * , 517 ( 1957 ) .
f. a. korolev , e. .n .
akimov , e. n. markov , and o. f. kulikov sov .
. dokl . * 1 * , 568 ( 1957 ) .
p. joos , phys .
* 4 * , 558 ( 1960 ) .
a. a. sokolov , n. p. klepikov and i. m. ternov , jetf * 23 * , 632 ( 1952 ) .
a. a. sokolov , and i. .m .
ternov , jetf * 25 * , 698 ( 1953 ) .
a. a. sokolov and i. .m .
ternov , _ synchrotron radiation _ , pergamon press , new york ( 1968 ) ; a. a. sokolov and i. .m .
ternov , _ radiation from relativistic electrons _ ,
translation series , new york ( 1986 ) .
i. m. ternov , physics - uspekhi * 38 * , 409 ( 1995 ) .
v. a. bordovitsyn , ph .
d. thesis , moscow ( 1983 ) ; i. m. ternov and v. a. bordovitsyn , vestn .
* 24 * , 69 ( 1983 ) ; v. a. bordovitsyn and v. v. telushkin , nucl .
inst . and meth .
* b266 * , 3708 ( 2008 ) .
s. a. belomesthnykh _ et al .
. inst . and meth .
* 227 * , 173 ( 1984 ) . j. le duff , p. c. marin , j. l. manson , and m. sommev , orsay - rapport technique , 4 - 73 ( 1973 )
. v. v. dmitrenko _ et al .
phys .- tech .
* 28 * , 1440 ( 1983 ) ; a. e. bolotnikov _ et al .
phys .- tech .
phys . * 33 * , 449 ( 1988 ) ; c. levin _ et al .
inst . and meth .
* a332 * , 206 ( 1993 ) . g. tepper and j. losee , nucl .
inst . and meth .
* a356 * , 339 ( 1995 ) .
e. nakamura _
et al . _ ,
j. of elec . spec . and
rel . phen . * 80 * , 421 ( 1996 ) ; d. e. baynham , p. t. m. clee , and d. j. thompson , nucl .
instr . and meth . ,
* 152 * , 31 ( 1978 ) k. sato , j. of synchrotron rad . , * 8 * , 378 ( 2001 ) ; t. gog , d. m. casa and i. kuzmenko , cmc - cat technical report . m. sands , slac technical note , slac-121 ( 1970 )
. b. norum , cebaf technical note , tn-0019 ( 1985 ) .
a. e. bolotnikov and b. ramsey , nucl .
inst . and meth .
* a396 * , 360 ( 1997 ) .
http://www.nist.gov/physlab/data/xraycoef/index.cfm http://library.lanl.gov/cgi-bin/getfile?00415886.pdf prajwal mohanmurthy prajwal mohanmurthy obtained his bachelor of science degree from mississippi state university in 2012 .
he was a graduate research fellow at the high performance computing collaboratory at mississippi state university and currently is a graduate research fellow in the laboratory for nuclear sciences at massachusetts institute of technology .
his research interests are centered around test of standard model and fundamental symmetries in search of physics beyond the standard model .
his recent research involvements have been geared towards a search for axionic dark matter and the precision measurement of the mass of neutrinos .
he also actively collaborates to develop beam instrumentation for up and coming facilities and future accelerators .
dipangkar dutta dr .
dipangkar dutta is an associate professor of physics at the mississippi state university department of physics and astronomy .
he obtained his bachelor of technology degree from indian institute of technology , bombay in 1992 and his doctoral degree in physics from northwestern university in 1999 .
he was a post - doctoral and senior post - doctoral fellow in the laboratory for nuclear sciences at massachusetts institute of technology .
his research is focused primarily on precision measurement of fundamental properties of nucleons .
he is also interested in precision tests of fundamental symmetries and the standard model . | the physics program at the upgraded jefferson lab ( jlab ) and the physics program envisioned for the proposed electron - ion collider ( eic ) include large efforts to search for interactions beyond the standard model ( sm ) using parity violation in electroweak interactions .
these experiments require precision electron polarimetry with an uncertainty of @xmath0 0.5 % .
the spin dependent synchrotron radiation ( sr ) , called spin - light , can be used to monitor the electron beam polarization . in this article
we develop a conceptual design for a `` spin - light '' polarimeter that can be used at a high intensity , multi - gev electron accelerator .
we have also built a geant4 based simulation for a prototype device and report some of the results from these simulations .
shell : bare demo of ieeetran.cls for journals polarized electrons , synchrotron radiation , spin light , differential ionization chambers . |
nitrogen - vacancy ( nv ) centers located deep within a diamond lattice appear promising as solid - state spin qubits since they combine optical initialization and readout capabilities @xcite , long electron spin coherence lifetimes ( approaching @xmath0 at room temperature @xcite ) , and the ability to control coupling to individual nuclear spins @xcite .
while these capabilities were initially demonstrated in high - purity natural diamond samples , similar results have recently been achieved in high - purity cvd - grown diamond @xcite .
there is now much interest in fabricating spin - based devices in diamond , with potential applications in quantum communication @xcite , quantum computation @xcite , and magnetometry @xcite . in all of these applications
it is necessary or at least advantageous to couple nv centers to optical structures such as waveguides @xcite and microresonators @xcite , to enable communication between distant qubits or to allow efficient extraction of emitted photons .
therefore , a reliable method is needed to create nv centers with good spectral properties in close proximity ( @xmath1 ) to a diamond surface .
in addition , the charge state of the nv center must be controlled ; all of the results mentioned above were demonstrated using nv centers in their negatively - charged state , nv@xmath2 .
it has been well established that nv centers can be created by ion implantation or irradiation followed by annealing @xcite .
ion implantation is used to create damage ( carbon vacancies ) in the diamond lattice , and it can also be used to introduce nitrogen impurities @xcite .
the choices of particle and acceleration voltage determine the initial vertical distribution of vacancies .
annealing is performed at temperatures ( typically @xmath3 ) where vacancies become mobile and can combine with nitrogen impurities to form nv centers .
vacancies can also recombine with self - interstitials @xcite ( implantation produces one interstitial carbon for each vacancy ) , can join together to form extended defects @xcite , or can become trapped at dislocations or at the surface @xcite . while numerous annealing experiments have been performed , and vacancy migration over long distances in the horizontal direction has been observed at @xmath4 annealing temperature @xcite , to our knowledge none of these studies has yet clearly answered the question of how far vacancies diffuse vertically from the initially damaged layer at temperatures optimized to create nv centers .
this is of vital importance for positioning nv centers close to an optical structure . here , we present etching experiments that reveal the vertical distribution of optically active nv centers produced using ion implantation and annealing to convert native nitrogen into nv centers .
the results indicate that vacancies can diffuse several hundred nanometers to form nv centers at the annealing temperatures used .
different behavior was observed for high - nitrogen ( @xmath5 ) and low - nitrogen ( @xmath6 ) diamond . in low - nitrogen diamond ,
spectroscopy results show an unexpected dependence of the nv@xmath2 signal on etch depth , and we interpret this behavior as an electronic depletion effect .
the first sample used in the experiments was a @xmath7 , @xmath8-oriented cvd diamond ( element 6 ) with a specified nitrogen concentration below @xmath9 , showing fairly uniform nv concentration before processing .
the sample was implanted with @xmath10 ga ions ( core systems ) with a dose of @xmath11 .
a monte - carlo simulation using srim @xcite predicts that this implantation energy produces vacancies to a depth of approximately @xmath12 , as shown in fig .
[ fig : implant ] . during implantation
, the cvd sample was covered with a tem grid to allow measurement of the photoluminescence ( pl ) intensity contrast between implanted and non - implanted surfaces .
this technique was necessary because the cvd sample has a high background nv concentration relative to its nitrogen content .
after implantation , the sample was annealed at @xmath13 in an h@xmath14/ar forming gas for 3 hours . the second sample was a @xmath15 , @xmath8 hpht diamond ( sumitomo ) with a specified nitrogen concentration of @xmath16 .
this sample has growth sectors with widely varying impurity concentrations . in most sectors , the width of the nv@xmath2 zero - phonon line ( zpl ) measured at low temperature is broad ( @xmath17 ) , suggesting a high nitrogen concentration , and the nv luminescence achievable by implantation and annealing in these sectors is very high .
however , as we have previously observed in sumitomo hpht samples @xcite , there exist particular sectors , with boundaries aligned to the sample edges , that have properties consistent with much lower nitrogen concentration , including a narrow zpl linewidth ( as low as @xmath18 ) at low temperature and an inability to produce strong nv luminescence with high implantation doses .
the spectral properties of this low - nitrogen sector in the hpht sample appear more similar to those of the cvd sample described above .
the hpht sample was implanted with @xmath10 ga ions ( core systems ) with a dose of @xmath19 .
a tem grid was not used on this sample , and was not required because the hpht sample has a very low background nv concentration relative to its nitrogen content . after implantation ,
the sample was annealed at @xmath13 in an h@xmath14/ar forming gas for 3 hours .
ga@xmath20 ion implantation.,width=336 ] , width=288 ] the samples were etched using an oxford inductively coupled plasma reactive ion etching ( icp - rie ) system with an o@xmath14/ar recipe @xcite . before etching , protective masks ( sio@xmath14 or sin )
were patterned on portions of the samples to allow measurement of the etch depth and for observation of the pl contrast between etched and non - etched regions .
after each trial , the etch depth was measured using a profilometer . since the step height seen by the profilometer includes both the diamond etch depth and the mask thickness , for some trials we also measured the etch depth by electron microscopy .
we measured the optical properties of the nv centers using a scanning confocal microscopy setup .
for excitation , a @xmath21 laser was focused through a 0.6 numerical - aperture microscope objective to a @xmath22 spot , and galvanometer mirrors or a piezo stage were used to scan the position of this spot on the sample .
the resulting pl was collected through the same objective and imaged onto a pinhole for spatial filtering , and detected using a cooled - ccd spectrometer or an avalanche photodiode with bandpass filters selecting wavelengths from approximately @xmath23 , for detection of the nv@xmath2 phonon sidebands .
the depth of focus was such that when the signal from a thin sheet of nv centers was measured as a function of the height of the microscope objective , a symmetric peak was observed with a full width at half maximum of @xmath24 , corresponding to @xmath25 in diamond .
the measurements were performed at room temperature . for both samples , images of the pl intensity
were obtained at the same locations after each trial . for the hpht sample
, spectra were also obtained at multiple locations , and from these spectra the intensities of the nv@xmath26 and nv@xmath2 zero - phonon lines at @xmath27 and @xmath28 , respectively , were individually determined .
[ fig : beforeimages ] shows pl images of the two samples before etching . for the cvd sample , the implantation pattern is visible , combined with ribbon - like features that must be caused by a non - uniform nitrogen concentration present in the sample before processing . for the hpht sample ,
the image shows a sharp boundary between growth sectors with different nitrogen concentrations .
the region that appears dark is the low - nitrogen sector mentioned above . from images similar to these
, the total photoluminescence intensity in the @xmath23 band was measured as a function of etch depth for the two samples .
the results , normalized by excitation power , are plotted in fig .
[ fig : apdplots ] . for the cvd sample ,
the quantity plotted is the average intensity measured in the implanted squares minus the average intensity measured outside of the squares . in this way we can separate the pl produced by nv centers created through ion implantation from the background nv pl . for the hpht sample ,
the initial nv concentration was very small , and only the total pl intensity is plotted in the figure , measured in both the high - nitrogen and low - nitrogen sectors . since the optical detection efficiency in our setup is approximately constant over @xmath29 of depth , the plotted quantities represent the total pl intensity from all of the remaining nv centers that were created through ion implantation and annealing . before presenting a detailed analysis , we note several important features of the data in fig .
[ fig : apdplots ] .
first , the behavior of the high - nitrogen sector of the hpht sample for the first @xmath30 of etching is quite different from that observed in the low - nitrogen sector of the hpht sample and in the cvd sample .
considering just the hpht sample , before etching , the total pl intensity is 44 times higher in the high - nitrogen sector than in the low - nitrogen sector , even through the implantation and annealing conditions were identical
. as the sample is etched , the pl intensity in the high - nitrogen sector rapidly decreases , indicating that most of the nv centers exist within the first @xmath31 , while the pl intensity in the low - nitrogen sector is approximately constant , suggesting that the first @xmath31 of material is devoid of optically active nv centers .
this is a strong indication that , for the high - nitrogen sector , the nv concentration produced through implantation and annealing was limited mainly by available vacancies , while for the low - nitrogen sector the nv concentration was limited by the available nitrogen .
however , at an etch depth of @xmath32 , the pl intensities in the two sectors are almost equal ( they differ by factor of 1.7 ) , and at this point the nv concentration is probably limited by the available vacancies in both sectors . as the etching progresses further ,
the weak background nv signals become important .
since the high - nitrogen sector has a higher background nv concentration , the pl intensities in the two sectors again diverge for the largest etch depths .
the pl measurement integrates contributions from all of the remaining nv centers produced by ion implantation and annealing , since the depth of focus of our optical setup is much larger than @xmath29 in diamond , as discussed above . to a first approximation ,
the nv concentration is therefore proportional to the first derivative of the pl intensity with respect to etch depth @xmath33 . however , there are two effects that complicate the analysis : spontaneous emission modification by the nearby diamond - air interface , and electronic depletion . spontaneous emission modification by a dielectric interface is well understood and can be included in the data analysis if we make an assumption about the radiative quantum efficiency of the nv center . in this section , we describe how to include spontaneous emission modification and obtain density estimates that are corrected for this effect .
electronic depletion is more complicated , but appears to be important only for the low - nitrogen material .
electronic depletion is discussed and modeled in the following sections .
the spontaneous emission rate of a quantum dipole is proportional to the total power radiated by an equivalent classical dipole . in the presence of a dielectric interface ,
the radiated power includes an interference term due to radiation reflecting off of the interface and performing work on the dipole .
we calculated the ratio between the total radiated power , including the dielectric interface , and that in bulk diamond , using a formula derived in ref . .
this ratio is equal to the spontaneous emission rate enhancement factor @xmath34 ( @xmath35 is the depth below the surface ) plotted in fig [ fig : nvdensity ] . for the nv center ,
we must include two dipoles that are orthogonal to the nv axis and to each other .
the nv axis can be oriented along any of the four @xmath36 crystal directions , and in our samples the surface is @xmath8 . for this geometry ,
the radiated power averaged over the two dipoles ( regardless of how they are chosen ) gives the same result as an isotropic mixture of dipole moments . for the result in fig .
[ fig : nvdensity](a ) , we have also averaged over @xmath37 emission wavelengths , to take into account the broad nv@xmath2 spectrum that is dominated by phonon sidebands . the spontaneous emission enhancement ( or suppression if @xmath38 ) is due entirely to the reflected wave , and only the radiated power into the substrate is modified .
the radiated power into air is unchanged , but as @xmath39 increases , the spontaneous emission lifetime of the nv center decreases , and thus the probability per excitation event for a photon to be emitted into air decreases . in the limit of weak excitation at @xmath21 , the rate of excitation events
does not depend on this spontaneous emission modification effect , and depends only on the driving electric field intensity , the dipole moment of the optical transition , and the spectral width of the phonon sidebands .
thus , in the weak excitation limit , assuming a radiative efficiency of unity , the collection efficiency is inversely proportional to @xmath39 . we have defined a theoretical , relative collection efficiency @xmath40 , plotted in fig .
[ fig : nvdensity](a ) .
we have ignored one subtlety by using a single @xmath41 based on the average radiated power from the two mutually orthogonal dipoles of the nv center . if there were no relaxation between excited states , then the two dipoles would each have a different spontaneous emission lifetime , and we would have to use @xmath42
. however , at room temperature , where rapid thermalization between the two orbital excited states may be expected , we believe it is more appropriate to average @xmath39 before taking the inverse . taking spontaneous emission modification into account
, we expect the measured intensity @xmath43 , where @xmath33 is the etch depth , to depend on @xmath44 and the initial nv concentration @xmath45 ( before etching ) according to , @xmath46 thus , to obtain @xmath45 we must solve a deconvolution problem . for our particular @xmath44
a converging iterative solution can be obtained as follows : @xmath47 where @xmath48
. the observed rate of convergence was such that , after 10 iterations , the remaining corrections had @xmath49 . in analyzing the data we must also contend with a large amount of noise .
we believe this noise is due mainly to errors in replacing the sample before each measurement and in setting the focus , and we also have uncertainty in estimating the etch depth .
the observed fluctuations are much larger than the poisson detection limit and are approximately 10% of the count rate .
we used the following procedure to estimate @xmath45 : ( 1 ) perform a linear interpolation of @xmath50 , and exponentiate to obtain a continuous version of @xmath45 .
( 2 ) calculate the corresponding @xmath45 using eqs .
[ eq : rho1]-[eq : rho3 ] .
( 3 ) smooth the result using a moving average with @xmath51 total width to obtain @xmath52 . ( 4 ) calculate a @xmath53 from @xmath52 using eq .
[ eq : sigma ] , and compare with the original data .
the calculated @xmath52 are plotted in fig .
[ fig : nvdensity ] , and the corresponding @xmath53 are the fits shown in fig .
[ fig : apdplots ] . including spontaneous emission modification had a noticeable effect on the calculated nv profiles , but the results are qualitatively similar to those obtained simply by differentiating @xmath43 . in order to estimate the absolute nv concentrations shown in the figure
, we also made use of separate calibration measurements performed on single nv centers , where the count rate per unit excitation power was approximately @xmath54 , and the effective collection area of the confocal setup was determined to be approximately @xmath55 .
the estimated sheet densities of optically active nv centers for the three measurements are @xmath56 and @xmath57 for the high- and low - nitrogen sectors , respectively , of the hpht sample , and @xmath58 for the cvd sample .
the results in fig . [ fig : nvdensity ] confirm that , for the high - nitrogen sector of the hpht sample , most of the optically active nv centers occur within the first @xmath31 .
the distribution is significantly broader than the predicted vacancy profile before annealing shown in fig .
[ fig : implant ] .
such a spreading of the distribution might be expected because of vacancy diffusion during annealing .
however , for the measurements on low - nitrogen material , the behavior is quite different . in these cases ,
the concentration of optically active nv centers is very low in the first @xmath12 , and the maximum concentration occurs approximately @xmath59 below the surface .
.,width=336 ] $ ] -oriented nv center close to a @xmath8 diamond surface .
( b - d ) nv concentration vs. depth , estimated from the data in fig . [ fig : apdplots].,width=288 ]
next , we present spectroscopy results from the hpht sample that help explain the lack of pl originating near the surface in the low - nitrogen material .
the insets of fig .
[ fig : spectra](a , b ) show typical room - temperature pl spectra from the high- and low - nitrogen sectors .
the most important features are the @xmath27 and @xmath28 peaks which have been identified as the zero - phonon lines ( zpls ) of nv@xmath26 and nv@xmath2 , respectively @xcite . the sharp line at @xmath60 that appears prominently in fig . [ fig : spectra](b ) is the diamond raman line .
to extract the nv@xmath2 and nv@xmath26 pl intensities from each spectrum , we fitted a voigt approximation @xcite to the @xmath27 and @xmath28 peaks after subtracting a linearly sloped background fitted to regions on the two sides of the peak .
the extracted peak areas are plotted as a function of etch depth in the main plots of figs .
[ fig : spectra](a , b ) . the nv@xmath26/nv@xmath2 intensity ratio for both sectors is plotted in fig . [ fig : spectra](c ) . and nv@xmath2 zero - phonon line ( zpl ) intensities vs. etch depth for the ( a ) high - nitrogen and ( b ) low - nitrogen sectors of the hpht sample .
insets : example spectra showing the diamond raman line at @xmath60 and nv@xmath26 and nv@xmath2 zero - phonon lines ( zpls ) at @xmath27 and @xmath28 , respectively .
( c ) the nv@xmath26/nv@xmath2 intensity ratio vs. etch depth for the high- and low - nitrogen sectors of the sumitomo sample.,width=288 ] for the first few hundred nanometers of etching , quite different behavior is observed for the high - nitrogen and low - nitrogen sectors . in the high - nitrogen sector , the nv@xmath26/nv@xmath2 ratio is approximately 0.2 and changes little as the sample is etched , although it eventually decreases .
however , for the low - nitrogen sector , this ratio changes dramatically as the sample is etched , starting out at 2.2 after the first successful etch , and decreasing to 0.3 after @xmath32 of etching .
even more surprisingly , for the first @xmath61 of etching the nv@xmath2 intensity actually increases .
the measured pl intensity includes the contributions from nv centers at all depths , and we do not expect significant absorption in the top layer .
we believe the most likely explanation is that the top layer interacts electronically with the layers below .
a simple model describing this interaction is presented in the next section .
finally , we briefly address the low - temperature spectral linewidth in the implanted samples , a critical issue for some potential applications .
in the ideal case , the linewidth of nv centers created through ion implantation and annealing would be equal to the linewidth observed in background nv centers distributed throughout the sample .
for another element 6 cvd sample with nitrogen content similar to that of the sample used in the etching experiment , the low - temperature linewidth ( fwhm ) of the nv@xmath2 zpl at @xmath62 was measured to be @xmath63 for the background nv centers . for the nv centers created by ion implanation
( @xmath64 , @xmath65 ga@xmath20 ) followed by etching to remove the top @xmath30 , the linewidth was @xmath66 , and the tails of the zpl peak were more prominent than for the background nv centers . for the hpht sample ,
the high - nitrogen sectors show broad linewidths ranging from @xmath67 .
in one of the special low - nitrogen sectors , a strain - split nv@xmath2 zpl with @xmath68 total linewidth was observed under normal excitation conditions .
however , when collecting pl through a gap waveguide using evanescent coupling @xcite a much broader linewidth ( @xmath69 ) along with a large nv@xmath26/nv@xmath2 ratio was observed at the same location .
this indicates that nv centers close to the surface had severely degraded properties .
we have performed low - temperature spectroscopy on many samples implanted by a variety of methods and plan to report the results elsewhere @xcite .
in this section we seek to explain the behavior in fig .
[ fig : spectra](b ) , especially the initial increase in the nv@xmath2 signal observed during the first @xmath61 of etching . in our view
the most likely explanation for this behavior is an electronic interaction , and in particular the presence of an electronic acceptor layer near the surface that removes electrons that would otherwise be available to form nv@xmath2 in the layers underneath .
it has been shown that the relative pl intensities of nv@xmath26 and nv@xmath2 depend on the concentrations of impurities such as nitrogen , and on the optical excitation conditions @xcite .
several types of impurities are present that can act as electronic donors and acceptors .
nitrogen impurities are expected to be the main electron donor , with an energy level @xmath70 below the conduction band minimum @xcite .
the nv center itself is known to exist in the nv@xmath2 and nv@xmath26 states , and perhaps could also exist in an nv@xmath20 state . because nv@xmath2 has discrete excited states @xmath71 above the ground states , the ground states of nv@xmath2 must be @xmath72 below the conduction band minimum @xcite . an abrupt change reported in the nv@xmath26/nv@xmath2 ratio at an excitation wavelength of @xmath73
@xcite suggests that the actual position of the nv@xmath2 ground state is @xmath74 below the conduction band minimum .
finally , we can only explain the results in fig .
[ fig : spectra ] if some species of electron acceptor is present in the top layer . as
discussed below , there are several possible acceptor species . for the present discussion we label the acceptor as a@xmath2 , and choose an energy level @xmath75 above the valence band maximum as expected either for substitutional ga impurities @xcite or for graphitic defects @xcite .
the indirect bandgap of diamond is @xmath76 .
theoretically , for a junction between a strongly `` p - type '' layer and an `` n - type '' layer ( nitrogen as the predominant impurity ) , a `` depletion layer '' of ionized nitrogen impurities is expected to have a width of approximately @xmath77e } = 90 \ , \mathrm{nm}$ ] if the nitrogen concentration is @xmath9 .
thus it is quite possible that a depletion layer could be responsible for a reduced nv@xmath2 concentration before etching in the low - nitrogen material . for a nitrogen concentration of @xmath78 ,
the theoretical depletion width is only @xmath79 , and interactions between different layers would be difficult to observe . as a first step toward quantitative modeling of the spectroscopy results ,
we have implemented a simple model that solves the 1-d poisson equation , @xmath80 + [ \mathrm{nv}^- ] + [ \mathrm{a}^- ] - p \right ) \
, , \label{eq : poisson}\ ] ] under the condition that the electronic occupations of each type of impurity , as well as the conduction - band electron concentration @xmath81 and valence - band hole concentration @xmath82 , are in thermal equilibrium according to fermi - dirac statistics .
. [ eq : poisson ] is then solved numerically under the constraint that the electric field is zero at @xmath83 , or equivalently , that the total charge is zero . in the actual experiment
, we do _ not _ necessarily expect the electronic populations to be in thermal equilibrium , since the sample is subject to optical excitation at @xmath21 .
nevertheless , given our lack of knowledge about all of the optical excitation rates , this model is a good starting point . in order to simulate the experiment
, we must also assume some distribution of nitrogen impurities , nv centers and acceptors .
as an example , we have simulated the following distribution , where @xmath84 is the depth below the original surface : @xmath85 & = & \left [ \mathrm{n}^0 \right ] + \left [ \mathrm{n}^+ \right ] = \mathrm{n}_\mathrm{bkg } - [ \mathrm{nv } ] \ , , \\ \left [ \mathrm{nv } \right ] & = & \left [ \mathrm{nv}^- \right ] + \left [ \mathrm{nv}^0 \right ] = \frac{r(z)}{1 + r(z ) } \mathrm{n}_\mathrm{bkg } \ , , \\
r(z ) & = & r_0 \left(1 + \theta(z - z_0 ) ( e^{-(z - z_0)/\delta z } - 1 ) \right ) \ , , \\
\left [ \mathrm{a } \right ] & = & \left [ \mathrm{a}^- \right ] + \left [ \mathrm{a}^0 \right ] = a_0 \theta(z_1-z ) \ , , \end{aligned}\ ] ] where @xmath86 is the initial nitrogen concentration ( assumed to be uniform ) , @xmath87 is a dimensionless rate for conversion of n to nv with a depth profile controlled by the parameters @xmath88 , @xmath89 and @xmath90 , @xmath91 is the unit step function , and @xmath92 is the concentration of acceptors in a top layer of thickness @xmath93 . we have chosen this impurity distribution because it is simple and yet contains the most essential features of the nv distribution and acceptor layer as discussed above . the simulations shown in fig . [ fig : simulations ] used the parameters @xmath94 ( @xmath95 ) , @xmath96 , @xmath97 , @xmath98 , and @xmath99 .
the temperature was chosen to be @xmath100 , which produces essentially the same results as @xmath101 .
[ fig : simulations](a ) shows the calculated 2-d concentrations of nv@xmath2 and nv@xmath26 ( the 3-d concentration integrated over @xmath35 ) as a function of etch depth . for these parameters ,
the simulation produces a similar behavior as observed in the spectroscopy data , where the nv@xmath2 pl intensity increases at first , and then decreases . to help in explaining this behavior ,
energy band diagrams and impurity electron concentrations are shown before etching ( b , c ) and after etching @xmath31 ( d , e ) . before etching
, the @xmath61 p - type layer takes electrons from nearby nv centers and nitrogen donors , so that nv centers can not exist in their negatively charged state to a depth of @xmath102 . below this depth ,
nv@xmath2 can exist , but the total nv concentration there is low .
however , after removal of the p - type layer , nearly all of the remaining nv centers can exist in their negatively charged state .
and nv@xmath2 populations ( integrated over @xmath35 ) vs. etch depth .
( b ) band diagram before etching , showing the conduction band minimum ( c.b . ) , valence band maximum ( v.b . ) , nitrogen donor level ( n@xmath26 ) , nv level ( nv@xmath2 ) , and acceptor level ( a@xmath2 ) , relative to the fermi level ( at zero energy ) .
( c ) populations before etching of neutral nitrogen impurities ( n@xmath26 ) , negatively charged nv centers ( nv@xmath2 ) , neutral nv centers ( nv@xmath26 ) , and acceptors ( a ) vs. depth @xmath35 .
( d ) band diagram after etching @xmath31 .
( e ) populations after etching @xmath31 . ,
width=316 ] thus , through a suitable choice of parameters , we can obtain good qualitative agreement with the spectroscopy results shown in fig .
[ fig : spectra](b ) for the low - nitrogen sector , demonstrating that electronic depletion is a plausible explanation for the initial rise in the nv@xmath2 signal .
such agreement is possible only for a limited range of parameters .
a substantial increase in nv@xmath2 due to removal of the p - type layer is possible only if @xmath86 is small , and @xmath86 can not be much larger than the value of @xmath95 used here .
this in turn limits the total number of nv centers that can exist in the model .
the maximum sheet density of nv@xmath2 obtained in the model , @xmath103 , is in good agreement with the experiment : as discussed above , we estimated @xmath57 . in order for the nv@xmath2 population to increase
gradually as the p - type layer is etched , the acceptor concentration must also be small , comparable to the initial nitrogen concentration in this case .
if it is much higher , the nv@xmath2 population remains approximately constant until the p - type layer is almost completely removed , at which point the nv@xmath2 population abruptly increases .
the acceptor concentration used here , approximately @xmath104 in a @xmath61 layer , corresponds to a sheet density of only @xmath105 , or only 7% of the implanted ga dose .
one important difference between theory and experiment can be seen in the nv@xmath26/nv@xmath2 ratio . in the model
, this ratio falls rapidly to zero once the damaged layer has been removed , and the nv concentration in the remaining layers is less than the nitrogen concentration . in the experiment ,
this ratio reaches a minimum of approximately 0.25 , and in general we observe this ratio to be dependent on the optical excitation power .
this is a limitation of the thermal equilibrium model presented above , especially at low temperature . as the temperature in the model
is increased ( a crude way to simulate intense optical excitation ) , the ratios will change , and the equilibrium value will eventually depend on the effective masses of electrons and holes in the conduction and valence bands . to model the system more accurately , we would need to use a rate - equation model for optical excitation combined with drift and diffusion of electrons and holes in the conduction and valance bands .
such a non - equilibrium model is feasible to solve numerically , but the main difficulty is that we have little knowledge of the ionization rates of nitrogen impurities , nv centers , and acceptors under optical excitation , as well as the electron and hole capture rates for these impurities , all of which are important . we have observed time - dependent behavior in the low - nitrogen sector that gives some hint of these rates : for excitation with @xmath106 at @xmath21 we observed that if the excitation is suddenly switched on , the pl intensity is initially smaller and increases by a factor of two with @xmath107 rise - time .
this rise time could be associated , for example , with optically induced charge transfer from acceptors to nv centers in the depleted region by means of free carriers .
finally , let us briefly address the question of what acceptor species could be responsible for creating such a depletion layer .
several possibilities exist .
damage close to the surface , created by polishing , for example , could have been present even before ion implantation .
if such damage consists of graphitic defects , this could pin the fermi level @xmath75 above the valence band maximum @xcite .
however , such defects would have to be distributed throughout the top @xmath108 of diamond , in a low concentration , to explain the observed result .
ion implantation introduces vacancies , self - interstitials , and ga impurities .
the subsequent annealing process is complex , and the final end - products are uncertain .
monovacancies could remain , but with an acceptor level apparently positioned @xmath109 above the valence band maximum @xcite ( nearly identical to nv@xmath2 ) , it is difficult to explain the observed results in terms of this acceptor . however , if vacancies combine to form extended defects @xcite , a high concentration of acceptor states throughout the diamond bandgap could exist .
finally , it is quite possible that the implanted ga atoms are the dominant acceptor .
initially we expected that ga impurities would not be important , since according to the simulation in fig .
[ fig : implant ] , @xmath110 vacancies are produced for each implanted ga@xmath20 ion .
however , the sheet density of acceptors needed to produce a depletion layer in our model is only 7% of the implanted ga@xmath20 dose .
acceptor levels for substitutional ga impurities and for ga - vacancy complexes have recently been calculated , with positions @xmath111 and @xmath112 above the valence band maximum , respectively @xcite .
from the data and simulations presented above , we can draw several conclusions related to formation of nv centers through implantation of heavy ions followed by annealing .
first , the vertical distance over which vacancies diffuse to form nv centers is only a few hundred nanometers , even for a fairly high annealing temperature of @xmath113 .
second , if the implantation dose is chosen to maximize the pl intensity in diamond with a moderately low nitrogen concentration ( @xmath114 ) , most of the nv@xmath2 pl that is produced originates from nv centers @xmath115 below the surface , while nv centers closer to the surface are either in the nv@xmath26 state or in other states that do not emit in our detection band .
such a situation is not useful for coupling nv centers to optical structures . finally , the spectroscopic results in the low - nitrogen diamond are highly suggestive of electronic depletion effects .
our model shows that the observed behavior can be explained by the presence of electronic acceptors near the surface , which could consist of impurities or defects introduced through the ion implantation and annealing , or damage originally present in the sample .
controlling such effects will be important in finding improved methods to couple dense ensembles of nv centers to optical structures on a diamond surface , and also for fabrication of single negatively - charged nv centers in ultra - pure diamond , where electronic depletion layers thicker than @xmath116 are possible . | etching experiments were performed that reveal the vertical distribution of optically active nitrogen - vacancy ( nv ) centers in diamond created in close proximity to a surface through ion implantation and annealing .
the nv distribution depends strongly on the native nitrogen concentration , and spectral measurements of the neutral and negatively - charged nv peaks give evidence for electron depletion effects in lower - nitrogen material .
the results are important for potential quantum information and magnetometer devices where nv centers must be created in close proximity to a surface for coupling to optical structures . |
during the past few decades , elaborate solar neutrino @xcite and atmospheric neutrino @xcite experiments have provided a wealth of convincing evidences for the existence of massive neutrinos and neutrino mixing , which could have an essential impact on particle physics , astrophysics and cosmology .
attentions have been focused on solving the puzzles of unexpected discrepancies between calculated and observed neutrino fluxes . instead of the more difficult and unlikely solution from an improved solar model @xcite , the solar @xmath3 deficit could be reconciled with the prediction if neutrino oscillations occur either in vacuum or in the presence of solar matter .
the flavor oscillation can be parameterized by the mass - squared differences of the neutrino mass eigenstates @xmath4 and @xmath5 , the mixing angles between weak eigenstates and mass eigenstates of the neutrinos ( @xmath6 is assumed ) . in terms of these parameters , the just - so vacuum oscillation @xcite requires @xmath7 ev@xmath8 and @xmath9 , while the msw resonant effect @xcite in the sun becomes important if @xmath10 ev@xmath8 @xmath11 ev@xmath8 , @xmath12 ( large angle solution ) , or
@xmath13 ev@xmath8 @xmath14 ev@xmath8 , @xmath15 ( small angle solution ) @xcite .
recent atmospheric neutrino data from the super - kamiokande @xcite further provide a strong evidence in support of neutrino oscillation as the cause to deficit of muon neutrinos , provided @xmath16 ev@xmath8 and @xmath17 .
it is clear that this solution to the neutrino anomaly in the atmosphere represents quite a distinct area in the parameter space as compared to that of the solar neutrino deficit .
based on the conclusive lep experiment at cern @xcite that there are three flavors of light , active neutrinos participating in the weak interaction , a direct - mass hierarchy @xmath0 , with @xmath18 ( @xmath19 and @xmath20 ) naturally accommodates the scales of both the two mass - squared differences and provides solutions to both puzzles : the conversion @xmath21 causes the observed deficit in the solar @xmath3 flux and the vacuum oscillation @xmath22 suppresses the @xmath23 flux in the atmosphere .
in addition to the sun and the atmosphere , type - ii supernovae are also natural sources that emit neutrinos . despite the first - ever observation of sn neutrino signals from sn 1987a @xcite , detailed neutrino spectral shapes
have not yet been determined with certainty due to low statistics and the physical processes that are not well understood .
this difficulty is accompanied by , for instance , the uncertainties in the characteristic temperatures @xmath24 as neutrinos were emitted from the neutrino - spheres .
consequently , the interpretation of future measurements of sn neutrinos would contain ambiguity in that the observed spectrum , which may have been deformed through conversion processes , could be simulated by different set of parameters at different temperatures .
it is therefore worthwhile to investigate how the uncertainty in @xmath24 could impact the interpretation of events at terrestrial detectors . in this paper , the parameters that solve solar and atmospheric neutrino problems
are taken as inputs , a natural choice as also adopted by some earlier works @xcite @xcite .
in addition , with the uncertainty in @xmath24 considered , we study whether a particular set of parameters could be singled out by future observations of sn neutrinos .
unlike solar neutrinos , the initial neutrino flux from a supernova contains all flavors of neutrino : @xmath25 and their anti - particles . under the direct - mass hierarchy of neutrinos
, the original @xmath2 spectra will be modified by the msw effect as neutrinos propagate through the resonance .
the @xmath26 spectra , on the contrary , is subject only to vacuum oscillation which yields an large averaged survival probability of @xmath27 .
the high - energy @xmath28 would not be converted to the easily detectable @xmath29 ( for instance , at super - kamiokande ) through the msw effect unless neutrino masses are inverted , in which case the heavier mass eigenstate has a larger component in @xmath3 than in @xmath23 or @xmath30 . since the mixing angles are defined in the first octant , the weak eigenstates @xmath31 , and @xmath3 are predominant in the mass eigenstates @xmath32 , and @xmath33 , respectively .
under the direct - mass hierarchy where @xmath34 , the mass eigenstates follow the hierarchy @xmath35 , while in the inverted - mass hierarchy , for instance , @xmath36 , the pattern @xmath37 follows . some models and phenomenological consequences involving inverted neutrino masses
have been discussed @xcite .
although current msw solutions to the solar neutrino problem ( snp ) have excluded @xmath3 as the heavier eigenstate , the inverse hierarchy could remain viable if the just - so vacuum oscillation is the solution for the snp .
if the inverted masses do apply and the resonance conditions for the anti - neutrinos are met , this could lead to an effective conversion between @xmath29 and the higher - energy @xmath38 to yield copious @xmath29-type events in the earth - bound detector . with the uncertainty in @xmath24 considered
, it is our second goal to investigate influences of both the direct and inverted mass hierarchies to future observations of sn neutrinos and how future measurements can play a role in this unsettled issue of direct versus inverted neutrino masses .
this paper is organized as follows . in section
ii we summarize the general features of stellar collapse and properties of the emitted neutrinos , and show how the uncertainty in neutrino temperature could affect the outcomes in the detector .
section iii and section iv contain more general results expected from the future observations at both super - kamiokande and sno , for direct and inverted masses , respectively .
based on the measurements , possible schemes which could provide discrimination among input parameters and between the two mass hierarchies are proposed .
section v contains discussions and our concluding remarks .
a massive star ( @xmath39 ) becomes unstable at the last stage of its evolution .
when the mass of the iron core reaches the chandrasekhar limits ( @xmath40 ) , it begins to collapse into a compact object of extremely high density , and the gravitational binding energy is released in the form of neutrinos .
et al . _
@xcite have pointed out that the total emitted energy , the averaged neutrino luminosity , and the mean neutrino energy are independent of the explosive mechanism but depend only on the mass of the initial iron core . regardless of the details of collapsed and bounce , it is well established that to form a typical neutron star after the collapse , an amount of @xmath41 erg , about 99% of the binding energy would be released in the form of neutrinos . each ( anti)neutrino species will carry away about the same amount of energy .
neutrinos are emitted from a collapsed star through two different processes : neutronization burst during the pre - bounce phase and thermal emission in the post - bounce phase .
the neutronization burst of a @xmath3 flux is produced by the electron capture on protons : @xmath42 .
the thermal emission creates @xmath43 pairs of all three flavors via the annihilation of @xmath44 pairs : @xmath45 .
the duration of neutronization burst lasts about a few millisecond and takes away 1%-10% of the total binding energy .
the thermal emission phase has a much wider spread of time structure , in the order of 10 seconds .
the initial neutrino spectrum is usually approximated by a fermi - dirac or a boltzmann distribution with a constant temperature and zero chemical potential .
to reduce the high - energy tail of the fermi - dirac distribution , some elaborate models introduce a nonzero chemical potential @xcite .
it is clear that the event numbers in a detector depend crucially on the @xmath2 temperature .
however , the numerical calculations based upon various models and physical arguments give rise to relatively wide ranges of temperature for each neutrino species @xcite and this uncertainty in @xmath24 could complicate the signatures concerning the oscillation of neutrinos from a supernova .
one may refer to ref .
@xcite for a review of neutrino oscillations .
although all the three flavors are emitted from a supernova , the phenomenon of sn neutrino oscillation can be well described through @xmath46 and @xmath47 @xcite . under the direct - mass hierarchy
, the probability @xmath48 is nearly independent of energy and is approximated by the vacuum oscillation expression .
the probability @xmath49 is energy - dependent and contains four parameters under a proper parameterization of the mixing matrix in the full 3-@xmath2 formalism : @xmath50 . in what follows , these four parameters
will simply be denoted as @xmath51 , and @xmath52 , respectively . among the above four parameters , the angle @xmath52 is special in certain aspect .
in addition to the limit @xmath53 at 90% c.l . set by the chooz @xcite long baseline reactor in the disappearance mode @xmath54 , an analysis in ref .
@xcite also has given allowed ranges of @xmath55 and @xmath56 for @xmath57 .
note that to zeroth order of @xmath58 , the probability becomes @xcite @xmath59 one may examine @xmath52 in more details through the iso - probability contours for sn neutrinos at several distinct scales of @xmath60 , as shown in figure 1 . within the interested range of @xmath61 , ( @xmath62 , or @xmath63 ) , the @xmath49 contours are almost independent of @xmath52 for @xmath64 ( @xmath65 ) . the parameter @xmath52 begins to show slight influence on @xmath49 contours only at very small @xmath61 and very large @xmath52 .
hence , for our purpose the choice of @xmath52 within @xmath57 would only affect the results slightly .
the value @xmath66 will be adopted for definiteness . as for other input parameters ,
the following are taken : large angle msw solution(la ) : @xmath67 ev@xmath68 small angle msw solution(sa ) : @xmath69 ev@xmath70 just - so vacuum solution(js ) : @xmath71 ev@xmath72 .
in what follows , an initial flux described by a fermi - dirac spectrum with zero chemical potential will be assumed .
the detailed time evolution during the cooling phase has been ignored , while the averaged magnitudes and effective temperatures of @xmath73 flux are used instead @xcite .
the event numbers at the detectors for the neutrino of type @xmath74 are estimated by @xmath75 here @xmath76 is the number of targets in the detector , @xmath77 is the initial number of @xmath78 , @xmath79 is the distance between the supernova and the earth , @xmath80(e@xmath81 ) is the cross section for the corresponding reaction , @xmath82 is the surviving probability for @xmath78 , @xmath83 is the temperature for @xmath78 , and @xmath84}.\ ] ] in evaluating the surviving probability , the electron number per nucleon is assumed to remain a constant ( @xmath85 ) and the density profile outside the neutrino - sphere ( @xmath86 cm ) is described by the power - law @xmath87 . to grasp the picture on how uncertainties in neutrino temperatures could affect the interpretation of observed events , we may tentatively assume @xmath88=3 mev , @xmath89= 6 mev ( @xmath90 ) , and compare outcomes from @xmath91=3 mev and @xmath91=4.5 mev . at super - kamiokande @xcite , contributions from the inverse beta decay @xmath92 predominate due to the high cross section .
events from other interactions : @xmath93 @xcite and @xmath94 @xcite , will also be included in our calculations although these events accumulate up to less than 5% of the @xmath95 events .
the threshold energy is taken to be 5 mev and the detector efficiency is assumed to be 100% . for 32 kton of water , one expects roughly @xmath96 neutrino events for a type ii supernova at the center of our galaxy ( @xmath97 10 kpc away ) .
since the cross section for @xmath98 is proportional to @xmath99 , and @xmath100 for the fermi - dirac distribution , a larger temperature gap between @xmath29 and @xmath101 would cause a more severely distorted spectra from the original one .
hence the difference between @xmath91 and @xmath102 determines to what extend the events are enhanced by oscillation . for the direct masses where @xmath0 , possible results of the ratio @xmath103 ( @xmath104 indicates oscillation , and @xmath105 indicates the case of no oscillation ) using specific input parameters
la , sa , and js are shown in figure 2 .
the curves representing la and sa are due to msw effects of the @xmath2-type events and the vacuum oscillation of the @xmath26-type events , while the js curve is due to vacuum oscillations of both @xmath2- and @xmath26-type events .
one observes that js parameters could raise event numbers most effectively , an increase of @xmath97 55% is possible at @xmath1063 mev ( @xmath107 ) .
the enhancement decreases as @xmath91 approaches @xmath102 . near a particular point
where @xmath108 , the conversion of @xmath29 to @xmath101 would not alter the original @xmath29 spectrum , all the scenarios yield @xmath109 and are indistinguishable among each other .
the complication arises from the fact that if , for instance , @xmath110 is observed , this observation is then either due to la parameters at @xmath111 ev or the js parameters at @xmath112 ev .
the uncertainties in neutrino temperatures would therefore render a wide range of predictions at the detectors .
informations such as the clues for oscillation and neutrino parameters would be hard to understand or even lost due to this complication .
for the observation of sn neutrinos , the extremely distinct time structures between the neutronization burst and the beginning of thermal emission would allow a clear separation at the @xmath113 cherenkov detector .
these two groups of events are discussed separately . the spectral shape and the total energy of @xmath3 from the early pre - bounce burst is still poorly known . for the purpose of qualitative discussion , the spectrum is arbitrarily chosen to be the same as that of thermal @xmath3 ( fermi - dirac ) with the same mean energy and a total of 5% the binding energy of a typical neutron star ( @xmath114 ) . during this early phase ,
one expects to observe the forward directional events due to elastic scattering @xmath115 and the backward events from @xmath116 .
these neutronization events are summarized in table i. we note that the forward events are relatively insensitive to the uncertainty in @xmath3 temperature .
the oscillation signature manifests itself through the drastically reduced forward events as compared to the original one , although practically the separation among la , sa , and js using events observed during this early phase is difficult . the backward events on the other hand , are more sensitive to @xmath117 .
the difficulty associated with the backward events comes from the extremely small numbers .
if @xmath118 mev and the total neutrino energy at this stage is down to @xmath119 of @xmath114 , the backward events are practically unobservable . because of the rapidly increased cross section for @xmath116 at higher energy : @xmath120 ,
the situation could be improved if @xmath117 is higher or if the neutrinos emitted during this phase have larger energy partition , which is quite model - dependent .
unlike the backward events , the forward event numbers are roughly in the order of 10 even if @xmath121 3 mev and the @xmath3 flux takes away as low as @xmath119 of @xmath114 . by using the numerous @xmath122 emitted from the inverse beta decay
, the distorted @xmath29 spectrum would be determined with better statistics . to account for the uncertainties in @xmath117 and @xmath123 , we may let @xmath124 @xmath91 , @xmath125 @xmath91 and compare the outcomes for @xmath91= 4 mev , 5 mev , and 6 mev .
the parameters @xmath126 and @xmath127 are allowed to vary within @xmath128 and @xmath129 to roughly include the temperature ranges given by current models .
expected ranges for the ratios @xmath130 are summarized in table ii , where @xmath131 includes events from the inverse beta decay and the neutrino interactions with oxygen , @xmath132 represents the forward scattering events . for @xmath133 mev
, there are two overlapped areas in @xmath134 : between no , sa and between js , la .
the same overlapping structure remains for @xmath135 mev and @xmath136 mev . despite the wealthy information conveyed through the @xmath122 spectrum at super - kamiokande , from table
ii it seems unlikely that a clear separation among input parameters could be achieved using the otherwise model - independent quantity @xmath134 unless the uncertainties in neutrino temperatures are reduced significantly .
the neutral current@xmath137 breakup reactions of deuterium in sno @xcite are flavor - blind for neutrinos : @xmath138 @xmath139 where @xmath140 .
the charged current reactions include two parts , @xmath141 @xmath142 with 1 kton of @xmath143 and a threshold energy @xmath97 5 mev ( 100% detection efficiency assumed ) , both @xmath144 and @xmath145 should roughly yield event numbers in the order of @xmath146 . the ratios @xmath147 at sno seem to provide a solution as to how a particular set of parameter could be singled out , as will be shown below . one may arbitrarily fix @xmath117 and parameterize other temperatures in a similar way :
let @xmath148 and allow an uncertainty in @xmath91 as well : @xmath149 , with @xmath150 and @xmath151 .
the ratio @xmath152 for @xmath153 and 5 mev are listed in table iii .
we found that even if the uncertainties in @xmath123 and @xmath91 may complicate the interpretation of observed events , each of the candidates gives rise to a distinct region in @xmath154 . in practical ,
if uncertainties in @xmath155 and @xmath156 can be reduced in the future , it would enable a smaller spread in each @xmath157 for a better separation .
in the light of msw effect , distinctions between direct and inverted masses would most likely appear in the observed neutrino spectra .
if the just - so vacuum oscillation is favored over the msw oscillations as solution to the snp , both direct ( @xmath0 ) and inverted - mass schemes ( @xmath1 ) are allowed since @xmath3 flux can also be converted to @xmath158 through the vacuum oscillation if neutrino masses are inverted .
the @xmath29 flux on the contrary , would go through the msw resonance if the mass hierarchy is inverted .
this conversion would presumably enlarge the @xmath29-type event rates effectively at super - kamiokande . without conflicting current solar and atmospheric neutrino data ,
we would focus on the js parameters : @xmath159ev@xmath8 and large @xmath61 , for a further investigation under the inverted mass scheme @xmath1 .
the possible sn @xmath29 spectra are shown in figure 3 .
curve @xmath160 is the original @xmath29 spectrum and curve @xmath161 represents the distorted one through the just - so vacuum oscillation under the direct - mass scheme while curve @xmath162 is obtained from the msw conversion under the inverted - mass scheme .
curves @xmath161 and @xmath162 nearly overlap , implying that the matter effect is not as prominent as expected , and that the msw effect under the inverted - mass hierarchy is almost identical to the vacuum oscillation under the direct - mass hierarchy for @xmath29 at this particular region of parameter space .
furthermore , the extremely small observable difference at the detectors would make the identification between the two mass patterns very difficult .
the reason becomes clear if the required conditions for a msw resonance and an adiabatic transition to occur are both considered @xcite : a density profile @xmath87 would yield @xmath163 ev@xmath8 relevant to the msw oscillation in the supernova .
this mass scale is much larger than the mass scale of js parameters ( @xmath164 ev@xmath8 ) .
therefore a very effective conversion of @xmath29 to @xmath101 in a supernova is unlikely for either direct or inverted masses if the js parameters are applied .
the strong conversion of @xmath29 to @xmath101 is actually disfavored by some analyses based on the sn 1987a data @xcite .
if either la or sa msw conversion is favored over the just - so vacuum oscillation , the case for the inverted hierarchy @xmath1 would then become shaky or can even be ruled out .
an alternative approach might shed some clues to the inverted - mass scheme and its outcomes
. we may characterize consequences for the oscillation @xmath165 by the surviving probability of @xmath29 in three limit cases : @xmath166 , and @xmath167 .
the case @xmath168 1 indicates that no conversion occurs among @xmath29 and @xmath101 , which is equivalent to the outcome of having massless neutrinos , and the mass pattern would then unlikely be the main issue .
the js parameters , as already discussed , yield @xmath169 for the msw conversion ( or equivalently , the vacuum oscillation under the direct - mass scheme ) .
one is therefore motivated to further study the consequences for a complete conversion of @xmath29 flux to @xmath101 , where @xmath170 . as pointed out by totani _
et al . _
@xcite , due to statistical uncertainties in experiments and the inconsistency among current analyses , one can not completely exclude the possibility of full conversion .
we may tentatively neglect details of the physical conditions and parameters that required for a complete conversion to occur , and assume that the probability @xmath48 remains approximately a constant within interested range of the neutrino energy .
to reasonably account for the contributions from @xmath3 and @xmath158 fluxes to the total events when a complete conversion occurs in the anti - neutrinos sector , one may first consider the contours of @xmath103 for events from the inverse beta decay only . under the inverted - mass scheme , a wide range of @xmath171 and @xmath61
are shown in figure 4 .
given @xmath172 4.5 mev , @xmath117=3 mev and @xmath173=6 mev , the js parameters ( @xmath164 ev@xmath8 and large @xmath61 ) roughly result in a 20% increment to the event number .
after a full conversion of the anti - neutrinos in which @xmath170 , one would expect to observe a sizable increase in the ratio @xmath103 .
therefore a larger @xmath174 and a smaller @xmath175 , at least several orders of magnitude , are required for a full conversion to occur .
the smallness of @xmath61 , alone with the small @xmath52 , fix the surviving probability of @xmath3 very close to unity .
hence , a full conversion of @xmath29 to @xmath101 would be accompanied by nearly unchanged @xmath3 and @xmath158 fluxes : @xmath176 .
the expected ratios @xmath177 at the super - kamiokande are listed in table iv . despite the better statistic provided by the @xmath29-type events at super - kamiokande ,
the detection is however not unique to the @xmath29 flux , the uncertainties in @xmath88 and @xmath178 therefore make the separation between @xmath170 ( complete conversion ) and @xmath179 difficult . at the sno detector ,
the charged - current channel @xmath180 is unique to @xmath29 .
this channel can be distinguished from the neutral - current events and the other charged - current channel induced by @xmath3 .
therefore the measurement of @xmath181 events at sno should be sensitive to the full conversion of @xmath29 to @xmath101 .
ratios of the neutral - current events to the charged - current events @xmath181 , denoted as @xmath182 , are shown in table v. we also present values of the same ratio under the direct - mass scheme in table vi as a comparison .
we observe that for a particular @xmath117 , both direct and inverted schemes yield a nearly identical range of @xmath182 if the js parameters are applied , this verifies a previous argument .
the valuable message from this ratio is that for a particular @xmath117 , a complete conversion of the @xmath29 flux through msw resonance represents a unique range of @xmath182 as compared to other scenarios , including that of the direct masses .
typical @xmath182 contours for inverted masses are shown in figure 5 , where @xmath117= 3 mev , @xmath91= 4.5 mev , and @xmath183= 6 mev are assumed . since the neutral - current reaction is blind to the oscillation , a complete swap of @xmath29 and @xmath101 fluxes would definitely yield smaller @xmath182 .
calculations show that @xmath184 for a complete conversion and indicate that @xmath185 ev@xmath8 and @xmath186 are required for a near complete conversion to occur .
tables v and vi suggest that for the detection of sn neutrinos , the direct and inverted masses could be distinguishable if a nearly complete conversion of @xmath29 to @xmath101 occurs , which yields a low @xmath182 and signals the existence of inverted - mass pattern .
since the msw effects become important for supernova neutrinos at @xmath187 ev@xmath8 , a future supernova would provide a test ground for @xmath188 ev@xmath8 .
if the nearly full conversion @xmath189 is observed in the sn neutrino flux , the consequences may have certain implications in that the required parameter spaces for a full conversion are obviously disfavored by current solar neutrino data , while the future solar and atmospheric observations may not severely change the mass scales required to explain the solar and the atmospheric neutrino deficits .
under the constraints of mass scale from solar and atmospheric neutrinos , the @xmath190 scenario naturally leads to four possible hierarchies ( one direct and three inverted ) : @xmath191 case 1 is the normal , direct mass scheme . for case 2 , @xmath174 in the mass scale of the msw solution
has been discussed @xcite . in our analysis
we have applied case 3 , in which the mass scale of @xmath174 is suitable for the vacuum solution of snp ( @xmath192 ev@xmath8 ) .
case 3 and case 2 become equivalent if @xmath164 ev@xmath8 since the msw and the vacuum oscillations for the @xmath29 flux would be nearly identical at this mass scale , as shown in section iv . for case 4 to survive , @xmath174 needs to be in the order of @xmath192 ev@xmath8 for the vacuum solution to apply .
therefore , consequences for case 4 and case 1 become equivalent in the detection of supernova neutrinos if @xmath164 ev@xmath8 . in this work , responses at both super - kamiokande and sno detectors to neutrino fluxes coming from a supernova are studied under the consideration of uncertainties in neutrino temperatures . in particular , some phenomenological consequences for direct - mass and inverted - mass patterns of neutrinos are compared .
we may summarize our results as follows .
( a ) uncertainties in neutrino temperatures can allow various interpretations of neutrino parameters .
we have shown this through the expected outcomes at super - kamiokande for sn neutrinos .
( b ) the three candidates la , sa , and js manifest differently in the ratio @xmath154 at sno even if the uncertainties in neutrino temperature are allowed .
future detection of sn neutrinos at sno would be able to single out favored mass and mixing parameters from the three candidates .
( c ) in addition to the direct - mass pattern , the inverted - mass scenario @xmath1 is investigated since it can allow the vacuum solution to the solar neutrino problem . by using the event ratio @xmath182 in sno , the direct - mass(@xmath0 ) and the inverted - mass(@xmath1 ) could be distinguished if a nearly complete @xmath189 conversion occurs in the anti - neutrino section .
s.c . would like to thank nien - po chen and sadek mansour for suggestions in preparing the manuscript .
t. k. is supported in part by the doe , grant no . de - fg02 - 91er40681 .
r. davis jr .
, prog . in nucl . and part
* 32 * ( 1994 ) ; sage collaboration , j. n. abdurashitov _
et al . _ ,
lett . b * 328 * , 234 ( 1994 ) ; gallex collaboration , p. anselmann _
et al . _ ,
lett . b * 357 * , 237 ( 1995 ) ; kamiokande collaboration , k. hirata _
et al . _ ,
d * 44 * , 2241 ( 1991 ) .
g. g. raffelt and j. silk , phys .
lett . b * 366 * , 429 ( 1996 ) ; d. o. caldwell and r. n. mohapatra , phys . lett . b * 354 * , 371 ( 1995 ) ; g. m. fuller , j. r. primack and y. -z .
qian , phys .
d * 52 * , 1288 ( 1995 ) . | possible outcomes of neutrino events at both super - kamiokande and sno for a type - ii supernova are analyzed considering the uncertainties in sn neutrino spectra ( temperature ) at emission , which may complicate the interpretation of the observed events . with the input of parameters deduced from the current solar and atmospheric experiments , consequences of direct - mass hierarchy @xmath0 and inverted - mass hierarchy @xmath1
are investigated .
even if the @xmath2 temperatures are not precisely known , we found that future experiments are likely to be able to separate the currently accepted solutions to the solar neutrino problem ( snp ) : large angle msw , small angle msw , and the vacuum oscillation , as well as to distinguish between the direct and inverted mass hierarchies of the neutrinos . |
an optical transmitter is employed to interrogate a distant region engulfed in bright thermal light , suspected of containing a weakly reflecting target .
the return light is processed by a receiver to decide whether or not the target is present .
recently @xcite , lloyd , building up on work by sacchi @xcite showed that in the above scenario , a quantum illumination " transmitter , i.e. , one that uses entangled light at the transmitter and an optimal quantum receiver , can perform substantially better than an un - entangled coherent laser transmitter , despite there being no entanglement left between the target - return and the idler beams due to high loss and noise .
this is the first example of an entanglement - based performance gain in the bosonic - channel setting where the initial entanglement does not survive the loss and noise in the system .
more recently @xcite , tan et .
al . showed that using a sequence of identical two - mode - squeezed gaussian states obtained from spontaneous parametric downconversion ( spdc ) , in conjunction with an optimal receiver , one may obtain up to a factor of 4 ( i.e. , 6 db ) gain in the error - probability exponent over an optimum - reception coherent - state radar in a highly lossy and noisy scenario .
this optimal receiver can be abstractly expressed as a projective measurement that projects onto the positive eigenspace of the difference of the density operators of the multi - mode states of the target return and retained idler modes , under the two hypotheses
@xmath0 : target absent , and @xmath1 : target present
. however , no structured receiver design is yet known that can harness any of the 6 db performance gain .
we present the design of a structured receiver , which despite being inferior to the ( yet un - implemented ) optimal joint - detection scheme , in conjunction with the spdc transmitter is shown to achieve up to a factor of 2 ( i.e. , 3 db ) error - exponent advantage over the optimum - reception classical sensor in the high loss , low brightness , high noise regime .
our receiver attempts to directly measure the remnant phase - sensitive cross - correlations between the return - idler mode pairs , by mixing the return and the idler beams on a parametric amplifier and photo - detecting the output .
numerical evidence suggests that our receiver achieves the asymptotic error - exponent of the optimal separable measurement , hence indicating that any superior receiver would have to make a complex joint measurement over multiple return - idler mode pairs .
an spdc transmitter generates @xmath2 independent spatio - temporal signal - idler mode pairs @xmath3 ; @xmath4 , each prepared in an identical entangled two - mode - squeezed state with a fock - basis representation @xmath5 where @xmath6 is the mean photon number in each signal and idler mode . in the quadrature representation , @xmath7 is a maximally - entangled zero - mean gaussian state with second - order quadrature field moments given by @xmath8 where @xmath9 , @xmath10 , and the standard quadrature field commutators @xmath11 = 0 $ ] and @xmath12 = [ { \hat a}_{i_m}^{(k ) } , { \hat a}_{i_n}^{(k ) } ] = { ( i/2)\delta_{mn}}$ ] apply for @xmath13 from the annihilation operators whenever convenient . as each signal - idler mode pair is prepared identically and undergo an identical channel transformation
, this will cause no ambiguity in notation . ] .
each signal mode is transmitted serially toward a spatial region that may or may not contain a weakly - reflecting specular point target , but in either case , contains a bright thermal - noise bath . under hypothesis @xmath0 ( no target ) , the target - return mode @xmath14 , where @xmath15 is in a thermal state with mean photon number @xmath16 . under hypothesis @xmath1 ( target present ) , @xmath17 , where the overall channel transmissivity @xmath18 , and @xmath15 is now in a thermal state with mean photon number @xmath19 . under @xmath0 the joint return - idler state for each of the @xmath2 mode pairs
@xmath20 is a product of two zero - mean thermal states @xmath21 with mean photon numbers @xmath22 and @xmath6 respectively . under @xmath1 , the joint return - idler state for each of the @xmath2 mode pairs
@xmath23 is a zero - mean gaussian state with second - order quadrature field moments given by @xmath24 the _ binary detection problem _ is to decide between hypotheses @xmath0 and @xmath1 ( assuming they are equally likely ) with minimum probability of error , using a quantum joint - detection measurement on the state of the @xmath25 return - idler mode pairs at the receiver , @xmath26 ; where @xmath27 or @xmath28 , given hypotheses @xmath0 or @xmath1 respectively .
helstrom derived the minimum probability of error @xmath29 for discriminating two quantum states as a function of the number of available copies ( or trials ) @xmath2 @xcite , @xmath30 , where @xmath31 are the non - negative eigenvalues of @xmath32 .
building up on the classical chernoff bound in classical detection theory @xcite , recently audenaert et .
al . derived the quantum chernoff bound ( qcb ) as an upper bound to @xmath29 , and also showed the qcb to yield the exact asymptotic rate exponent of the minimum error probability @xcite . combining that with a ( relatively loose ) lower bound on @xmath29 @xcite , and defining @xmath33 and @xmath34
, we have @xmath35 where the second inequality ( qcb ) is asymptotically tight as @xmath36 .
the qcb is customarily represented as @xmath37 in terms of an error - rate exponent @xmath38 .
the third inequality is a loose upper bound known as the bhattacharyya bound .
symplectic decomposition of gaussian - state covariance matrices was used to compute the qcb explicitly @xcite , for both the coherent - state and the entangled ( spdc ) transmitters @xcite , and it was shown that in the high loss , weak transmission and bright background regime , i.e. , with @xmath39 , @xmath18 , and @xmath16 , the entangled transmitter yields a qcb error - exponent @xmath40 , which is four times ( or @xmath41 db ) higher than the error - exponent @xmath42 for a coherent - state transmitter with identical per - mode average transmitted photons as the entangled transmitter . in fig .
[ fig : bounds ] , we plot the regions captured by the upper and lower bounds in for both the classical and the entangled transmitters , showing a clear advantage of quantum over classical illumination . transmitted modes each with @xmath43 mean transmitted photons per - mode , @xmath44 mean thermal - noise photons per mode , and @xmath45 . the coherent - state transmitter lower bound also applies to all classical - state transmitters with a total of @xmath46 photons in all @xmath2 transmitted modes ; thus depicting an undisputed asymptotic advantage obtained by the quantum illumination ( spdc ) transmitter with joint optimum reception over all @xmath2 return - idler mode pairs .
the curve plotted with ( magenta ) circles depicts the error - probability performance of the coherent - state transmitter and mode - by - mode homodyne - detection receiver , which is asymptotically optimal ( as @xmath36 ) for the coherent - state transmitter in the @xmath16 regime .
the thick black line is the performance of the opa receiver that we propose in this paper , which performs substantially better compared to the optimal classical sensor ( magenta circles ) note that the magenta circles is the best error - probability performance that can be achieved by _ any _ classical transmitter ( the error - probability lower bound of the blue region is rather loose and anything below the magenta circles can not be achieved by any classical transmitter ) .
the error - probability performance of the ( yet unknown ) optimum joint - detection receiver for the spdc transmitter will lie in the ( red / light shaded ) region between the bounds for the quantum - illumination transmitter , and will hug the quantum - illumination transmitter upper bound for high values of k.,width=302 ]
for a coherent - state transmitter , each received mode @xmath47 is in a thermal state with mean photon number @xmath22 , and a mean - field value @xmath48 or @xmath49 for hypotheses
@xmath0 and @xmath1 respectively . hence ,
homodyne detection on each received mode @xmath50 yields a variance-@xmath51 gaussian - distributed random variable @xmath52 with mean @xmath53 or @xmath49 given the hypothesis .
assuming both hypotheses to be equally likely , the minimum error probability rule to decide between @xmath0 and @xmath1 is to use the sufficient statistic @xmath54 against a threshold detector , i.e. , say @xmath55 if @xmath56 and @xmath57 otherwise . the probability of error is given by the gaussian error - function : @xmath58 where the approximation holds for @xmath59 and @xmath60 is the error - exponent .
for @xmath16 , @xmath61 , identical to the qcb error - exponent for the coherent - state transmitter .
it is straightforward to see that for @xmath2 large enough , @xmath62 .
therefore in the high - background regime , mode - by - mode homodyne detection is asymptotically optimal for the coherent - state transmitter .
for the spdc transmitter , each received return - idler mode pair @xmath63 is in a joint gaussian state with zero mean under both hypotheses , and covariance matrix @xmath64^{t}\left [ \begin{array}{cccc } { \hat a}_r^{\dagger } & { \hat a}_i^\dagger & { \hat a}_r & { \hat a}_i \end{array } \right]\rangle$ ] whose entries are readily computed using the quadrature field moments in eqs . , and .
for hypothesis @xmath1 , @xmath65 \nonumber } } \ ] ] and for hypothesis @xmath0 , @xmath66 . unlike in the coherent - state transmitter case , the entangled transmitter results in zero - mean joint return - idler states under both hypotheses .
as is evident from the covariance matrices under @xmath0 and @xmath1 , in the @xmath67 , @xmath18 , @xmath16 regime , the sole distinguishing factor between the two hypotheses that makes quantum illumination perform superior to the un - entangled coherent - state transmitter , are the off - diagonal terms of @xmath68 bearing the remnant phase - sensitive cross - correlations of the return - idler mode pairs when the target is present , @xmath69 .
the optimal joint - detection receiver acts on all the @xmath25 received return - idler modes and yields the minimum probability of error @xmath29 by capturing the remnant return - idler phase - sensitive cross - correlations for the @xmath1-hypothesis in the most efficient way .
return - idler mode pairs pairwise on an optical parametric amplifier ( opa ) with gain @xmath70 , and counts the total number of clicks @xmath71 on a photon counter at one output port of the opa over all @xmath2 output modes .
the receiver decides in favor of hypotheses @xmath0 or @xmath1 depending upon whether @xmath72 or @xmath73 , @xmath74 being the decision threshold.,width=340 ] the receiver approach we propose uses an optical parametric amplifier ( opa ) which is a type - ii degenerate amplifier constructed from a @xmath75 non - linear crystal .
the incident return and idler modes @xmath76 and @xmath77 are combined and amplified by an opa driven by a strong local pump beam , producing pairs of output modes @xmath78 where @xmath79 is the gain of the opa ( see fig . [
fig : oparcvr ] ) .
each output mode @xmath80 is in a zero - mean thermal state with mean photon number under the two hypotheses given by @xmath81 i.e. , @xmath82 , for @xmath83 for @xmath0 and @xmath1 respectively .
hence , the joint state of the @xmath2 received modes @xmath84 is a @xmath2-fold tensor product @xmath85 of identical zero - mean thermal states with per - mode mean photon number @xmath86 or @xmath87 depending upon which of the two hypothesis is true .
a @xmath2-fold product of thermal states is diagonal in the @xmath2-fold tensor - product of photon - number bases of the @xmath2 modes .
hence the optimum joint quantum measurement to distinguish between the two hypotheses is to count photons on each output mode @xmath84 and decide between the two hypotheses based on the total photon count @xmath71 over all @xmath2 detected modes , using a threshold detector .
the probability mass function of @xmath71 under the two hypotheses is given by @xmath88 where @xmath89 or @xmath28 based on which hypothesis is true .
the mean and variance of this distribution are @xmath90 and @xmath91 respectively , where @xmath92 . for large @xmath2 the above conditional distributions for @xmath71 approach gaussian distributions
( due to the central limit theorem ( clt ) ) with means and variances given by @xmath90 and @xmath91 respectively .
the probability of error is given by @xmath93 where an error - exponent @xmath94 can be achieved using a threshold detector that decides in favor of hypotheses @xmath0 or @xmath1 depending upon whether @xmath72 or @xmath73 , with @xmath95 .
@xmath96 is a function of the opa gain @xmath70 .
given @xmath39 , @xmath18 , @xmath16 , intuitively a small gain , @xmath97 with @xmath98 will be optimal to distinguish between the two hypotheses , such that the difference in the total mean photon count is dominated by the term proportional to the phase - sensitive correlation , i.e. , @xmath99 , for @xmath39 , @xmath18 , @xmath16 and @xmath100 .
for the problem parameters as chosen in fig .
[ fig : bounds ] , i.e. , @xmath43 , @xmath45 and @xmath44 , @xmath96 is maximized for @xmath101 , confirming our intuition that a small gain is optimal . for this gain , @xmath102 . for the above parameters , error - exponents for the classical and spdc transmitters with optimal measurement are @xmath103 and @xmath104 respectively .
hence for these parameters , our receiver gets @xmath105 db gain in the error - exponent over the classical system . finally , to establish the asymptotic error - exponent performance of the opa receiver , we will use the classical bhattacharyya bound to the error - probability @xmath106 .
the bhattacharyya bound to the error probability in distinguishing between the two distributions @xmath107 and @xmath108 using k i.i.d .
observations is given by : @xmath109 where @xmath110 using the approximation @xmath111 , we obtain @xmath112^{-1 } \\ & = & \frac{1}{1 + \frac{\delta^2}{8n_0(1+n_0 ) } } \\ & \approx & 1 - \frac{(n_1-n_0)^2}{8n_0(1+n_0 ) } \\ & = & 1 - r_b,\end{aligned}\ ] ] with @xmath113 .
this translates to the bound @xmath114 where the bhattacharyya bound exponent @xmath115 is given by @xmath116 for a choice of @xmath117 , for @xmath39 , @xmath18 , @xmath16 ( @xmath100 is automatically satisfied ) satisfying @xmath118 would work as well . ] . as @xmath119 ,
therefore by construction , for a weak transmitter operating in a highly lossy and noisy regime , the opa receiver achieves at least a @xmath120 db gain in error exponent over the optimum - receiver classical sensor whose qcb error exponent @xmath121 . for @xmath39 and @xmath100 , both @xmath86 and @xmath122 .
hence , a single - photon detector ( as opposed to a full photon - counting measurement ) suffices to achieve the performance of the receiver depicted in fig .
[ fig : oparcvr ] . , where @xmath31 are the non - negative eigenvalues of @xmath123 , using clt to compute the error - probability with independent measurements on all return - idler mode pairs , followed by majority - vote hard decision.,width=340 ]
entanglement has proven to be one of the most useful resources in quantum information , with applications to improving classical communication rates over noisy quantum channels , teleportation of unknown quantum states over long distances , and in quantum algorithms that can potentially solve certain problems ( such as factoring ) faster than the corresponding best known conventional classical algorithms . whereas realizing useful quantum computing is not practically feasible with current technology , exploiting quantum effects to build optical communications and sensing systems that could perform better than corresponding classical systems seems to be well within the reach of current state - of - the - art in experimental quantum optics .
quantum illumination " is a novel preliminary attempt at using an entangled source of light to detect the presence of a weakly - reflecting target in the far - field subject to a very lossy and noisy channel @xcite .
even though the efficacy of using spontaneous parametric downconversion a well - known entanglement source as a transmitter to obtain a significant advantage ( 6 db in error - exponent ) over using a classical laser transmitter for target detection was established by means of quantum chernoff bounds @xcite , a structured receiver design that may harness this advantage is not known to date .
our work puts forth an explicit receiver design using conventional quantum optics , that when used along with an spdc transmitter is capable of obtaining up to a 3 db error - exponent gain over the optimum - reception classical sensor .
there are several problems and challenges both in theory and experiments that remain open .
it is still not clear whether a sequence of identically prepared two - mode - squeezed states constitutes the best quantum - illumination transmitter .
it will hence be of interest to explore whether a transmitter that uses non - gaussian entangled states , or complex higher - order entanglement between all the transmitted signal - idler mode pairs can achieve more than just a constant - factor improvement in the error exponent .
it would also be interesting to extend this work to detecting and imaging spatially extended speckled targets . in terms of receiver design ,
the problem of constructing a receiver for the spdc transmitter that achieves the full 6 db error - exponent gain is certainly at the forefront .
numerical evidence suggests that any alternative receiver that makes independent measurements on each return - idler mode pair is not likely to perform significantly better than the opa receiver in the absence of classical feedforward or soft - decision post - processing through the @xmath2 measurement instances improvement in the error - exponent as the opa receiver . ] , thus necessitating a superior receiver to make a joint measurement over _ all _ return - idler mode pairs ( see fig .
[ fig : helstrom_compare ] ) .
the current model assumes the receiver to have complete knowledge of the signal power , channel loss and noise power for the optimal design of the receiver .
whereas the knowledge of the transmitted signal power at the receiver is a reasonable assumption , a full estimation - theoretic study of the performance of quantum illumination with no prior knowledge of channel loss and noise , would be necessary in order to design a realistic prototype of the quantum sensor .
the author thanks jeffrey h. shapiro , mit for valuable discussions and for pointing out an error in the second - order moment calculations .
the author also thanks zachary dutton , bbn , seth lloyd , mit and baris i. erkmen , jpl for several interesting discussions on this topic and for feedback on this draft .
the author thanks the darpa quantum sensors program and bbn technologies corporation for supporting this research .
in order to substantiate our claim about the performance of single - shot helstrom measurement ( depicted in fig . [
fig : helstrom_compare ] ) , we compute the error - probability performance of the pair - wise ( separable ) hard - decision helstrom measurement ( the quantum measurement that minimizes the single - shot probability of error ) followed by a majority - vote detector after independently detecting all @xmath2 return - idler mode pairs . in order to compute the helstrom measurement minimum probability of error @xmath124 , we need to compute the non - negative eigenvalues of the difference of the density operators of the two - mode return - idler states under the two hypotheses , i.e. @xmath123 . for our propagation and noise model , these two return - idler density operators can be computed in full generality in terms of their respective tensor - product fock - state basis elements . for a pure - state two - mode entangled transmitter of the form
@xmath125 the return - idler state matrix - elements come out to be : @xmath126 and @xmath127 , \nonumber\end{aligned}\ ] ] when @xmath128 is satisfied and @xmath129 for @xmath130 .
@xmath131 and @xmath132 for @xmath133 are the photon - number states for the return and the idler mode respectively , and each form a complete orthonormal set of bases for the respective state spaces .
@xmath134 is the well - known hypergeometric function . using the above expressions for the density operators
, we computed the error - probability performance of the symbol - by - symbol ( separable ) helstrom minimum error - probability measurement on each return - idler mode pair , and numerical results seem to suggest that this measurement has the same error - exponent as ( and performs only slightly better than ) the opa receiver .
the separable helstrom measurement performance was calculated by evaluating the single - shot minimum error - probability @xmath135 , where @xmath31 are the non - negative eigenvalues of @xmath123 , and then using the central limit theorem to compute the error - probability with independent measurements on all return - idler mode pairs , followed by a majority - vote hard decision ( see fig .
[ fig : helstrom_compare ] ) .
10 s. lloyd , science * 321 * , 1463 ( 2008 ) .
tan , _ et .
* 101 * , 253601 ( 2008 ) .
s. pirandola and s. lloyd , phys .
a * 78 * , 012331 ( 2008 ) .
m. f. sacchi , phys .
a * 71 * , 062340 ( 2005 ) ; m. f. sacchi , phys .
a * 72 * , 014305 ( 2005 ) . c. w. helstrom , _ quantum detection and estimation theory _ , vol .
123 , academic press , new york ( 1976 ) .
h. chernoff , ann .
* 23 * , 493 ( 1952 ) .
k. m. r. audenaert , _ et . al .
lett . * 98 * , 160501 ( 2007 ) ; j. calsamiglia , _ et .
a * 77 * , 012331 ( 2008 ) . | an optical transmitter that uses entangled light generated by spontaneous parametric downconversion ( spdc ) , in conjunction with an optimal quantum - optical receiver ( whose implementation is not yet known ) is in principle capable of obtaining up to a 6 db gain in the error - probability exponent over the optimum - reception un - entangled coherent - state lidar to detect the presence of a far - away target subject to entanglement - breaking loss and noise in the free - space link @xcite .
we present an explicit design of a structured quantum - illumination receiver , which in conjunction with the spdc transmitter is shown to achieve up to a 3 db error - exponent advantage over the classical sensor .
apart from being fairly feasible for a proof - of - principle demonstration , this is to our knowledge the first structured design of a quantum - optical sensor for target detection that outperforms the comparable best classical lidar sensor appreciably in a low - brightness , lossy and noisy operating regime . |
_ polynomial chaos _ ( pc ) is a computational spectral method that is used to propagate and quantify uncertainties arising in the modeling of physical systems .
the method originated from the works of wiener @xcite and cameron and martin @xcite on the decomposition of functionals in a basis of hermite polynomials of gaussian random variables .
ghanem and spanos @xcite used such an expansion to solve stochastic equations with random data .
xiu and karniadakis @xcite introduced _ generalized polynomial chaos _ ( gpc ) involving non - gaussian random parameters .
extensions to arbitrary probability measures followed in the works @xcite ; see also @xcite for recent theoretical developments and a convergence result for gpc expansions . _ polynomial chaos expansions _ ( pce )
provide an explicit expression of quantities of interests as functionals of the underlying uncertain parameters and in some situations allow us to perform uncertainty quantifications at a considerably lower computational cost than monte carlo methods .
however , they suffer from the curse of dimensionality and thus typically work efficiently for systems involving low dimensional uncertainties .
a related major drawback appears in the long - term integration of evolution equations . in the presence of random forcing in time ,
the number of stochastic variables in the system increases linearly with time and thus quickly becomes overwhelming .
moreover , standard pce utilize orthogonal polynomials of the initial distribution and as time evolves , the dynamics deviate from the initial data substantially ( e.g. , due to nonlinearities ) and the solution may become poorly represented in the initial basis .
this led the authors in @xcite to legitimately question the usefulness of ( standard ) pce to address long - term evolution properties of solutions such as intermittent instabilities .
this paper aims to address the aforementioned drawbacks .
we propose a pc - based method that constructs evolving chaos expansions based on polynomials of projections of the time dependent solution and the random forcing .
more precisely , chaos expansions at each iteration are constructed based on the knowledge of the moments of underlying distributions at a given time . in the setting of dynamics satisfying a markov property
, we crucially exploit this feature to introduce projections of the solution at prescribed time steps that allow us to `` forget '' about the past and as a consequence keep the dimension of the random variables fixed and independent of time .
our chaos basis is adapted to the evolving dynamics with the following consequences : ( i ) our expansion retains its optimality for long times ; and ( ii ) the curse of dimensionality is mitigated .
notably , we establish , with appropriate modifications , theoretical convergence analysis which sheds light on our numerical findings .
further , asymptotic analysis for computational complexity will also be discussed .
inspired by examples of @xcite , we will apply our method to a nonlinear coupled system of stochastic differential equations ( sdes ) and its variants .
other iterative methodologies have already been proposed in @xcite in various contexts of random differential equations ( rdes ) .
the main novelty of our approach is that it allows us to compute long - term solutions of evolution equations with complex stochastic forcing , which here will be a brownian motion but could be generalized to more complex models .
white noise driven evolution equations have important roles in modeling the small scale effects and certain uncertainties in many applications such as turbulence , filtering , mathematical finance , and stochastic control @xcite .
the plan of the paper is as follows .
section [ sec : pc ] introduces background material and establishes the necessary notation for pce .
our methodology , called dynamical generalized polynomial chaos ( dgpc ) , is described in detail in section [ sec : method ] .
numerical experiments comparing dgpc to hermite pc and monte carlo simulations are presented in section [ sec : numeric ] .
some conclusions are offered in section [ sec : conclu ] .
we briefly introduce the original pc framework . the notation we follow
can be found in @xcite . given a probability space @xmath0 where @xmath1 is a general sample space , @xmath2 is a @xmath3-algebra of subsets and @xmath4 is a probability measure , we consider @xmath5 , the space of real - valued random variables with finite second moments .
let @xmath6 be a countable collection of independent and identically distributed ( i.i.d ) standard gaussian random variables belonging to the probability space , and @xmath7 .
then , we define the wick polynomials @xmath8 where @xmath9 belongs to set of multi - indices with a finite number of nonzero components @xmath10 @xmath11 is the @xmath12th order normalized one - dimensional hermite polynomial and @xmath13 .
note that the wick polynomials are orthogonal to each other with respect to the measure induced by @xmath14 .
the cameron and martin theorem @xcite establishes that the wick polynomials form a complete orthonormal basis in @xmath15 .
this means that any functional @xmath16 , can be expanded as @xmath17,\end{aligned}\ ] ] and the sum converges in @xmath18 .
we define @xmath19 as expectation with respect to @xmath4 .
the expansion is called polynomial chaos expansion ( pce ) .
we will also use the term hermite pce to emphasize that the expansion utilizes gaussian random variables .
one of the notable features of is that it separates the randomness in @xmath20 such that the coefficients @xmath21 are deterministic and all statistical information is contained in the coefficients . in particular , the first two moments are given by @xmath22 = u_0 $ ] and @xmath23 = \sum_{\alpha \in \mathcal{j } } |u_{\alpha}|^2 .
$ ] higher order moments may then be computed using either triple products of wick polynomials @xcite or the hermite product formula @xcite ; see section [ sec : moments ] . in numerical computations
, the doubly infinite expansion is truncated so that it becomes a finite expansion @xmath24 where we simply used polynomials up to degree @xmath25 in the variables @xmath26 . throughout the paper ,
we use the graded lexicographic ordering for multi - indices ; see ( * ? ? ?
* table 5.2 ) . in the context of white noise driven sdes , the random variables @xmath27 can be obtained by the projection @xmath28 where @xmath29 is a brownian motion on @xmath30 $ ] and @xmath31 is a complete orthonormal system in @xmath32 $ ] .
moreover , @xmath33 is comprised of i.i.d .
standard gaussian random variables , and the expansion @xmath34 converges in @xmath18 for all @xmath35 and has the error bound @xmath36 ^ 2 = o(t k^{-1 } ) , \end{aligned}\ ] ] provided the @xmath31 s are trigonometric polynomials ; see and @xcite . here , brownian motion @xmath37 is projected onto @xmath32 $ ] for a fixed time @xmath38 so that the corresponding wick polynomials @xmath39 depend implicitly on @xmath40 .
the simple truncation leads to @xmath41 terms in the approximation .
thus , the computational cost increases rapidly with high dimensionality , which in turn decreases the efficiency of pce .
this is the `` curse of dimensionality '' .
another related major problem of pc is that expansions may converge slowly and even fail to converge for long time evolutions @xcite .
these problems led to numerous extensions of the hermite pce , which we now briefly discuss .
the paper @xcite proposed a method called generalized polynomial chaos ( gpc ) , in which the above random variables @xmath42 have specific , non - gaussian , distributions in the askey family @xcite .
further generalizations to arbitrary probability measures beyond the askey family were proposed in @xcite , where the probability space is decomposed into multiple elements and chaos expansions are employed in each subelement .
the approach taken in @xcite is to use a restart procedure , where the chaos expansion is restarted at different time - steps in order to mitigate the long - term integration issues of chaos expansions .
another generalization toward arbitrary distributions is presented in @xcite , using only moment information of the involved distribution with a data - driven approach .
our key idea is to adapt the pce to the dynamics of the system . consider for concreteness the following sde : @xmath43,\end{aligned}\ ] ] where @xmath44 is a general function of @xmath20 ( and possibly other deterministic or stochastic parameters ) , @xmath29 is a brownian motion , @xmath45 is a constant , and @xmath46 is an initial condition with a prescribed probability density .
we assume that a solution exists and is unique on @xmath30 $ ] .
as the system evolves , a fixed polynomial chaos basis adapted to @xmath46 and @xmath47 may not be optimal to represent the solution @xmath48 for long times .
moreover , the dimension of the representation of @xmath47 increases linearly with time @xmath40 , which renders the pce method computationally intractable even for moderate values of @xmath40 .
we will therefore introduce an increasing sequence of restart times @xmath49 and construct a new pce basis at each @xmath50 based on the solution @xmath51 and all additional random variables that need to be accounted for . a very similar methodology with @xmath52 was considered earlier in @xcite .
when random forcing is present and satisfies the markov property as in the above example , the restarting strategy allows us to `` forget '' past random variables that are no longer necessary and focus on a significantly smaller subset of random variables that influence the future evolution . as an example of application
, we will show that the restarting strategy allows us to capture the invariant measure of the above sde , when such a measure exists . to simplify notation
, we present our algorithm on the above scalar sde with @xmath44 a function of @xmath20 , knowing that all results also apply to systems of sdes with minor modifications ; see sections [ sec : higherdim ] and [ sec : numeric ] .
first , we notice that the solution @xmath48 of is a random process depending on the initial condition and the paths of brownian motion up to time @xmath40 : @xmath53 therefore , recalling the expansion for @xmath54 , the solution at time @xmath40 can be seen as a nonlinear functional of @xmath46 and the countably infinite variables @xmath14 .
as previously noticed in @xcite , the solution @xmath48 can be represented as a linear chaos expansion in terms of the polynomials in itself and therefore , for sufficiently small later times @xmath55 , @xmath56 , the solution @xmath57 can be efficiently captured by low order chaos expansions in @xmath48 everything else being constant .
moreover , the solution @xmath57 on the interval @xmath58 depends on @xmath59}$ ] and not on values of @xmath60 outside of this interval .
this significantly reduces the number of random variables in @xmath14 that need to be accounted for .
this crucial observation clearly hinges upon the markovian property of the dynamics .
hence , dividing the time horizon into small pieces and iteratively employing pce offer a possible way to alleviate both curse of dimensionality and long - term integration problems .
we decompose the time horizon @xmath30 $ ] into @xmath12 subintervals ; namely @xmath61,[t_1,t_2],\ldots,[t_{n-1},t]$ ] where @xmath62 .
the idea is then to employ polynomial chaos expansion in each subinterval and restart the approximation depending on the distributions of both @xmath63 and @xmath51 at each @xmath50 , @xmath64 ; see figure [ fig : prop_chaos ] . here
, @xmath63 denotes the gaussian random variables required for brownian forcing on the interval @xmath65 $ ] . throughout the paper ,
we utilize the term @xmath66 to represent orthonormal chaos basis involving its arguments . in order to establish evolution equations for chaos basis @xmath67 , triple products @xmath68
$ ] are needed .
this procedure basically corresponds to computing the coefficients and indices in the hermite product formula .
the probability distribution of @xmath51 does not belong to any classical family such as the askey family in general .
the construction of the orthogonal polynomials in @xmath51 is therefore fairly involved computationally and is based on the general pce results mentioned earlier in the section . at each time step , we compute the moments of @xmath51 using its previously obtained chaos representation and incorporate them in a modified gram schmidt method .
this is a computationally expensive step .
armed with the orthogonal basis , we then compute the triple products in @xmath51 to perform the necessary galerkin projections onto spaces spanned by this orthogonal basis and thus obtain evolution equations for the deterministic expansion coefficients . letting @xmath69 ,
equations for the coefficients @xmath70 of @xmath71 are given in the general form @xmath72 ds + \sigma \ , { \mathbb{e}}\left[t_{\alpha}({\boldsymbol{\xi}}_j , u_j ) \
, d{w_s}\right ] , \end{aligned}\ ] ] where @xmath73 . before integration of these expansion coefficients in time
, @xmath74 is represented in terms of its own orthogonal polynomials , which provides a description of the initial conditions on that interval .
we then perform a high order time integration .
cumulants of the resulting solution are then computed to obtain relevant statistical information about the underlying distribution
. the proposed methodology does not require the computation of any probability density function ( pdf ) and rather depends only on its moments at each restart time .
the statistical information contained in such moments provides useful data for decision making in modeling and , for determinate distributions , moments characterize the distribution uniquely .
this aspect will be essential in the proof of convergence in section [ sec : conv ] . comparing our method with probability distribution function methods , e.g. , the fokker
planck equation , we note that it essentially evolves the coefficients of orthogonal polynomials of the projected solution in time instead of evolving the pdf .
let @xmath75 represents the nonlinear chaos expansion mapping @xmath76 to @xmath77 .
our scheme may be demonstrated by figure [ fig : prop_chaos ] .
mathematically , we have @xmath78 in simulations , @xmath63 is truncated with finite dimension @xmath79 .
let also @xmath80 denote the maximum degree of the polynomials in the variables @xmath81 . here , to simplify
, @xmath82 and @xmath25 are chosen independently of @xmath83 .
the multi - indices having @xmath82 dimensions and degree @xmath25 belong to the set @xmath84 we now present our algorithm , called _ dynamical generalized polynomial chaos _ ( dgpc ) , in one dimension for concreteness ; see also section [ sec : higherdim ] . + * algorithm 1d * : dgpc + - decompose time domain @xmath30 = [ t_0,t_1 ] \cup \ldots.\cup [ t_{n-1},t]$ ] + - initialize degrees of freedom @xmath85 + - compute coefficients / indices in hermite product formula for @xmath86 + - integration in time : + - time step @xmath87 : obtain random variable @xmath74 + - calculate moments @xmath88 $ ] + - construct orthogonal polynomials @xmath89 + - compute triple products @xmath90 $ ] + - perform galerkin projection onto @xmath91 + - set up initial conditions for the coefficients @xmath92 + - evolve the expansion coefficients @xmath92 + - calculate cumulants + - postprocessing + several remarks are in order .
first , since our stochastic forcing has identically distributed independent increments , we observe that at each subinterval the distribution of @xmath63 is the same . computing and storing ( in sparse format ) the coefficients for the product formula only once drastically reduces the computational time . at each iteration
, the random variable @xmath93 is projected onto a finite dimensional chaos space and the next step is initialized with the projected random variable .
for a @xmath94-dimensional sde system , this projection leads to @xmath95 terms in the basis for the subinterval @xmath96 $ ] .
therefore , the total number of degrees of freedom used in @xmath30 $ ] becomes @xmath97 ; see also section [ sec : cost ] .
we emphasize that small values of @xmath85 are utilized in each subinterval such that computations can be carried out quickly .
the idea of iteratively constructing chaos expansions for arbitrary measures was considered earlier in @xcite . in @xcite ,
an iterative method is proposed to solve coupled multiphysics problems , where a dimension reduction technique is used to exchange information between iterations while allowing the construction of pc in terms of arbitrary ( compactly supported ) distributions at each iteration .
a similar iterative procedure for hermite pc was suggested for stochastic partial differential equations ( spdes ) with brownian forcing in the conclusion section of @xcite without any mathematical construction or numerical example .
we also stress an important difference between our approach and @xcite : our scheme solely depends on the statistical information , i.e. , moments , which are available through chaos expansion , whereas @xcite either requires the pdf of @xmath74 at @xmath50 or uses mappings to transform @xmath74 back to the original input random variables , which in our problem is high dimensional , including the @xmath42 variables .
we now describe the implementation of our algorithm .
provided that the distribution of the initial condition @xmath46 is known , several methods allow us to compute moments : analytic formulas , quadrature methods , or monte carlo sampling .
also , in the case of limited data , moments can be generated from raw data sets . throughout the paper , we assume that the moments of the initial condition can be computed to some large finite order so that the algorithm can be initialized .
we first provide the hermite product formula , which establishes the multiplication of two pce for random variables @xmath20 and @xmath98 .
the chaos expansion for the product becomes @xmath99 with @xmath100^{1/2 } , $ ] @xcite .
as our applications include nonlinear terms , this formula will be used repeatedly to multiply hermite chaos expansions . to compute moments of the random variable @xmath74 ( recall ) at time step @xmath50 , @xmath101
, we use its chaos representation and previously computed triple products recursively as follows . due to the markov property of the solution ,
the normalized chaos basis in @xmath102 and @xmath103 becomes a tensor product , and thus we can write @xmath104 where @xmath105 and @xmath106 . then for an integer @xmath107 , @xmath108 here , @xmath109 and @xmath110 . hence , using , the coefficients @xmath111 , where @xmath112 and @xmath113 , can be calculated using the expression @xmath114 \ , ( u_j^{m-1})_{\gamma-\theta+\alpha , l}\,(u_j)_{\theta+\alpha , k},\end{aligned}\ ] ] which allows us to compute moments by simply taking the first coefficient , i.e. @xmath115 = ( u_j^m)_{\boldsymbol{0},0}$ ] . even though the coefficients @xmath116 can be computed offline , the triple products @xmath117 $ ] must be computed at every step as the distribution of @xmath77 evolves . given the real random variable @xmath74 obtained by the orthogonal projection of the solution @xmath20 onto homogeneous chaos at @xmath50 , the gram
schmidt orthogonalization procedure can be used to construct the associated orthogonal polynomials of its continuous distribution . for theoretical reasons ,
we assume that the moment problem for @xmath74 is uniquely solvable ; i.e. , the distribution is nondegenerate and uniquely determined by its moments @xcite . to construct orthogonal polynomials , the classical gram
schmidt procedure uses the following recursive relation @xmath118 where the coefficients are given by @xmath119 \left({\mathbb{e}}[t^2_l(u_j)]\right)^{-1}$ ] .
note that @xmath19 denotes the lebesgue
stieltjes integral with respect to distribution of @xmath74 .
we observe that the coefficient @xmath120 requires moments up to degree @xmath121 as @xmath122 is at most of degree @xmath123 .
thus , normalization would need first @xmath124 moments for orthonormal polynomials up to degree @xmath125 .
after the construction of orthonormal polynomials of @xmath74 , the basis @xmath126 can be generated by tensor products since @xmath127 and @xmath74 are independent .
moreover , triple products @xmath128 $ ] , where @xmath129 can be found by simply noting that this expectation is a triple sum involving moments up to order @xmath130 .
hence , the knowledge of moments yields not only the orthonormal set of polynomials but also the triple products required at each iterative step . in our numerical computations , we prefer to utilize a modified gram
schmidt algorithm as the classical approach is overly sensitive to roundoff errors in cases of high dimensionality @xcite .
we refer the reader to @xcite for other algorithms , e.g. , stieltjes and modified chebyshev methods .
when the algorithm is reinitialized at the restart point @xmath50 , it needs to construct the initial condition in terms of chaos variables for the next iteration on @xmath65 $ ] . in other words
, we need to find the coefficients @xmath131 in terms of @xmath74 and @xmath127 . to this end
, we observe that the first two polynomials in @xmath74 are given by @xmath132 and @xmath133 ) , $ ] where @xmath134 is the standard deviation of the random variable @xmath74 .
notice that @xmath135 , because we assumed in the preceding section that the distribution was nondegenerate .
hence , the initial condition can be easily found by the relation @xmath136 t_0(u_j ) + \sigma_{u_j } t_1(u_j ) = \sum_{\alpha \in \mathcal{j } , \ , k \in \mathbb{n}_0 } ( u_j)_{\alpha , k } \ , t_{\alpha}({\boldsymbol{\xi}}_{j } ) \,t_k(u_{j}).\end{aligned}\ ] ] we now extend our algorithm to higher dimensions , and for concreteness , we consider two dimensions .
let @xmath137 be a two - dimensional random variable obtained by projection of the solution @xmath20 onto homogeneous chaos space at time @xmath138 .
the two - dimensional case requires the calculation of the mixed moments @xmath139 , l , m \in \mathbb{n } \cup \{0 \}$ ] . in the case of independent components
@xmath140 , the moments become products of marginal moments and the basis @xmath141 is obtained by tensor products of the one - dimensional bases . for a nonlinear system of equations ,
the components of the solution typically do not remain independent as time evolves .
therefore , we need to extend the procedure to correlated components . denoting by @xmath142 and @xmath143 the corresponding chaos expansions on @xmath65
$ ] so that @xmath144 and @xmath145 , we compute the mixed moments by the change of variables formula @xmath146= { \mathbb{e}}_{(v_1,v_2)}[v^l_1 \ , v^m_2 ] = { \mathbb{e}}_{({\boldsymbol{\xi}}_{j},u_{j } ) } \left[z_{j+1,1}^l \ , z_{j+1,2}^m \right],\end{aligned}\ ] ] where @xmath147 represents the expectation with respect to the joint distribution induced by the subscript .
note that the latter integral in can be computed with the help of previously computed triple products of @xmath63 and @xmath77 .
incidentally , similar transformation methods were used previously in @xcite .
since the components may not be independent , the tensor product structure is lost but the construction of orthogonal polynomials is still possible .
based on the knowledge of marginal moments , we first compute the tensor product @xmath141 . however , this set is not orthogonal with respect to the joint probability measure of @xmath148 . therefore , we further orthogonalize it via a gram schmidt method .
note that this procedure requires only mixed moments @xmath149 $ ] , which have already been computed in the previous step of the algorithm .
it is worth noting that the resulting set of orthogonal polynomials is not unique as it depends on the ordering of monomials @xcite . in applications
, we consider the same ordering used for the set of multi - indices .
finally , calculation of triple products and initial conditions can be extended in an obvious way . in this section ,
we discuss the computational costs of our algorithm using a vector - valued version of the sde in @xmath150 for a fixed @xmath151 . for each scalar sde
, we assume that the nonlinearity is proportional to @xmath125th order monomial with @xmath152 .
furthermore , we compare costs of dgpc and hermite pc in the case where both methods attain a similar order of error . throughout this section , @xmath85 will denote the dimension and the degree in @xmath153 for hermite pc .
here @xmath14 represents the truncation of @xmath94-dimensional brownian motion ; therefore , @xmath154 .
the time interval is fixed and given by @xmath155 $ ] .
we also let @xmath156 @xmath157 be the degrees of freedom for the simple truncation so that the dimension of truncated multi - indices @xmath158 becomes @xmath159 . for the dgpc method , we divide the interval into @xmath160 identical subintervals so that @xmath161 .
we now slightly change the notation of previous sections and denote by @xmath162 the dimension and degree of polynomials of @xmath163 used in dgpc approximation for each @xmath64 . with additional @xmath94 variables at each restart ,
the dimension of the multi - index set for dgpc becomes @xmath164 due to the tensor product structure .
further , let @xmath165 and @xmath166 denote the time step and global order of convergence for the time integration method employed in hermite pc , respectively . with the above notation , we summarize the computational costs in table [ table : compcost ] .
all terms should be understood in big @xmath167 notation .
r | p6.5cm|p3.9 cm flops & dgpc & hermite pc + offline & @xmath168 & @xmath169 + time evolution & @xmath170 $ ] & @xmath171 $ ] + moments & @xmath172 $ ] & + gram schmidt & @xmath173 & + triple products & @xmath174 & + initials & @xmath175 & + the estimates in table [ table : compcost ] are obtained as follows .
triple products for @xmath14 with @xmath176 can be calculated by the equation @xmath177 & = { \mathbb{e}}\left [ \prod_{i=1}^k h_{\alpha_i } ( \xi_i ) \prod_{j=1}^k h_{\beta_j } ( \xi_j ) \prod_{k=1}^k h_{\gamma_k } ( \xi_k ) \right]= \prod_{i=1}^k { \mathbb{e}}[h_{\alpha_i}(\xi_i)h_{\beta_i}(\xi_i)h_{\gamma_i}(\xi_i)].\end{aligned}\ ] ] thus , the offline cost is of order @xmath178 assuming that one - dimensional triple products can be read from a precomputed table . since a time discretization with error @xmath179
is employed for hermite pc , then for dgpc in each subinterval , time steps of order @xmath180 should be used to attain the same global error . without nonlinearity , the total cost in hermite pc for evolution of a @xmath94-dimensional system becomes @xmath181 .
due to the functional @xmath182 , the cost should also include the computation of @xmath183 at each time integration step .
this requires additional @xmath184 work ( see the computation of moments below ) .
hence , the corresponding total cost of evolution becomes the sum of these costs .
the coefficients of the @xmath185th power of a single projected variable @xmath74 are given by @xmath186 , \quad 2 \leq k \leq 3n_*,\ ] ] assuming that , after each multiplication , the variable is projected onto the pc basis .
here , @xmath187 .
thus , the computation of marginal moments of all variables requires @xmath188 calculations at each restart time in dgpc .
since we use the same ordering for multidimensional moments as @xmath189 , computing joint moments further needs @xmath190 amount of work . the gram
schmidt procedure costs are cubic in the size of the gram matrix , which has dimension @xmath191 in dgpc .
each triple product @xmath192 , \alpha,\beta,\gamma \in \mathcal{j}_{d , n_*}$ ] , is a triple sum ; therefore , the total cost is proportional to @xmath193 .
also , initial conditions can be computed by orthogonal projection onto the dgpc basis , which costs @xmath164 per dimension .
note that although we employed simple truncation for multi - indices , sparse truncation techniques can be utilized to further reduce the costs of both computing moments and evolution ; see @xcite .
we now discuss the error terms in both algorithms .
we posit that the hermite pce converges algebraically in @xmath25 and @xmath82 and that the error at time @xmath40 satisfies @xmath194 and @xmath195 depending on the sde . here
@xmath196 denotes that @xmath197 for @xmath198 with @xmath199 .
this assumption enforces an increase in the degrees of freedom @xmath200 if one wants to maintain the accuracy in the long term .
algebraic convergence in @xmath82 and the term @xmath201 stems from the error bound .
we do not show errors resulting from time discretization in since we already fixed those errors to the same order in both methods .
incidentally , even though we assumed algebraic convergence in @xmath25 , exponential numerical convergence , depending on the regularity of the solution in @xmath14 , is usually observed in the literature @xcite .
although it is usually hard to quantify the constants @xmath202 , this may be done for simple cases .
for instance , for the ornstein uhlenbeck ( ou ) process , using the exact solution and fourier basis of cosines , we can obtain the parameters @xmath203 and @xmath204 for short times .
the parameter @xmath205 does not play a role in this case since the sde stays gaussian and hence we take @xmath206 .
we also note that for complex dynamics , @xmath205 may depend on @xmath25 ; see @xcite in the case of the stochastic burgers equation and advection equations .
now , we fix the error terms in to the same order @xmath207 , @xmath56 , by choosing @xmath208 .
since dgpc uses truncated @xmath14 of dimension @xmath209 and polynomials of degree @xmath210 in each subinterval of size @xmath211 , the @xmath18 error at @xmath212 has the form @xmath213 of accuracy , we can choose @xmath214 from table [ table : compcost ] , we observe that to minimize costs for dgpc , we can take @xmath215 and @xmath216 and maximize @xmath12 according to the previous equation as @xmath217 with these choices of parameters , the total computational cost for dgpc is of algebraic order in @xmath82 and in general , is dominated by the computation of moments . for hermite pc with large @xmath82 and @xmath218 ( or @xmath219 in the case of exponential convergence in @xmath25 ) , both offline and evolution stages include the cubic term @xmath220 .
both costs increase exponentially in @xmath82 .
therefore , the asymptotic analysis suggests that dgpc can be performed with substantially lower computational costs using frequent restarts ( equation ) in nonlinear systems driven by high dimensional random forcing .
we now consider the convergence properties of our scheme as degrees of freedom tend to infinity .
there is an extensive literature on the moment problem for probability distributions ; see @xcite and their references .
we provide the relevant background in the analysis of dgpc . _
( hamburger moment problem ) _ for a probability measure @xmath221 on @xmath222 , the moment problem is uniquely solvable provided moments of all orders @xmath223 , @xmath224 exist and they uniquely determine the measure . in this case , the distribution @xmath221 is called _
there are several sufficient conditions for a one - dimensional distribution to be determinate in hamburger sense : these are compact support , exponential integrability and carleman s moment condition @xcite .
for instance , gaussian and uniform distributions are determinate , whereas the lognormal distribution is not .
the moment problem is intrinsically related to the density of associated orthogonal polynomials .
indeed , if the cumulative distribution @xmath225 of a random variable @xmath20 is determinate , then the corresponding orthogonal polynomials constitute a dense set in @xmath226 and therefore also in @xmath227 @xcite .
additionally , finite dimensional distributions on @xmath150 with compact support are determinate ; see @xcite .
now , let @xmath228 denote an independent collection of general random variables , where each @xmath229 ( not necessarily identically distributed ) has finite moments of all orders and its cumulative distribution function is continuous . under these assumptions ,
@xcite proves the following theorem .
[ thm : conv_ernst ] for any random variable @xmath230 , gpc converges to @xmath205 in @xmath18 if and only if the moment problem for each @xmath229 , @xmath231 , is uniquely solvable .
this result generalizes the convergence of hermite pce to general random variables whose laws are determinate .
notably , to prove @xmath18 convergence of chaos expansions , it is enough to check one of the determinacy conditions mentioned above for each one - dimensional @xmath229 .
we now consider the relation between determinacy and distributions of sdes .
consider the following diffusion @xmath232,\end{aligned}\ ] ] where @xmath54 is @xmath94-dimensional brownian motion , @xmath233 , and @xmath234 are globally lipschitz and satisfy the usual linear growth bound .
these conditions imply that the sde has a unique strong solution with continuous paths @xcite .
determinancy of the distribution of @xmath235 is established by the following theorem ; see @xcite for details .
[ thm : det_sde ] the law of the solution of is determinate if @xmath236 is finite . consider the solution of the sde written as @xmath237 where @xmath238 is the exact evolution operator mapping the initial condition @xmath46 and @xmath239 to the solution @xmath235 at time @xmath240 .
here , with a slight change of notation , @xmath241 represents brownian motion on the interval @xmath242 $ ] .
note that future solutions @xmath243 , @xmath244 , are obtained by @xmath245 , where @xmath246 denotes brownian motion on the interval @xmath247 $ ] .
even though the distribution of the exact solution @xmath235 is determinate under the hypotheses of thm .
[ thm : det_sde ] , it is not clear that this feature still holds for the projected random variables . to address this issue , we introduce , for @xmath248 , a truncation function @xmath249 @xmath250 where @xmath251 \subset { \mathbb{r}}^d$ ] , and such that @xmath252 decays smoothly and sufficiently slowly on @xmath253 such that the lipschitz constant @xmath254 of @xmath255 equals @xmath256 ; i.e. , for @xmath257 , the euclidean distance in @xmath150 , @xmath258 . such a function is seen to exist by appropriate mollification for each coordinate using the continuous , piecewise smooth , and odd function @xmath259 defined by @xmath260 on @xmath261 $ ] , and extended by @xmath262 outside @xmath263 $ ] .
this truncation is a theoretical tool that allows us to ensure that all distributions with support properly restricted on compact intervals remain determinate ; see @xcite .
we can now introduce the following approximate solution operators for @xmath248 and @xmath264 : @xmath265 where @xmath266 denotes the orthogonal projection onto polynomials in @xmath20 and @xmath241 , and @xmath159 is the total number of degrees of freedom used in the expansion . throughout this section ,
we use a linear indexing to represent polynomials in both @xmath20 ( restricted on @xmath267 ) and the random variables @xmath14 .
the main assumptions we imposed on the solution operator @xmath268 are summarized as follows : a. @xmath269 , where @xmath270 and @xmath271 with @xmath272 .
[ item : zero ] b. @xmath273 , where @xmath274 , @xmath275 , and @xmath271 with @xmath272 .
[ item : one ] c. for @xmath56 and @xmath276 , there is @xmath277 so that @xmath278 , where @xmath279 . [
item : two ] assumption [ item : zero ] is a stability estimate ensuring that the chain remains bounded in @xmath280th norm starting from a point in @xmath150 ( here @xmath262 without loss of generality owing to [ item : one ] ) .
assumption [ item : one ] is a lipschitz growth condition controlled by a constant @xmath281 .
these assumptions involve the @xmath282 norm , and hence are slightly stronger than control in the @xmath18 sense .
note that @xmath283 also satisfies assumption [ item : one ] with the same constant @xmath281 since @xmath284 .
finally , assumption [ item : two ] is justified by the weierstrass approximation theorem for the @xmath20 variable in @xmath267 and by the cameron martin theorem to handle the @xmath14 variable .
consider now the following versions of the markov chains : @xmath285 where @xmath286 ( possibly random and independent of the variables @xmath14 ) and @xmath287 is a sequence of i.i.d .
gaussian random variables representing brownian motion on the interval @xmath288 $ ] .
we also use the notation @xmath289 to denote brownian motion on the interval @xmath290 $ ] .
then we have the first result .
[ lemma : rapprox ] assume [ item : zero ] , [ item : one ] , and [ item : two ] hold , and let @xmath291 be finite .
suppose also that @xmath292 .
then , for each @xmath56 , there exists @xmath248 depending on @xmath12 , @xmath293 , and @xmath294 such that @xmath295 for each @xmath296 .
let @xmath56 be fixed .
using properties [ item : zero ] and [ item : one ] , we observe that @xmath297 where for all @xmath83 , @xmath127 and @xmath241 are identically distributed . then , by induction , we obtain @xmath298 where the constant @xmath299 is bounded .
the last inequality indicates that @xmath280th norm of the solution grows with @xmath83 , possibly exponentially , but remains bounded . for @xmath300
, we note that @xmath301 since @xmath74 is independent of future forcing . by markov s inequality and ,
we get @xmath302 let @xmath303 be the indicator function of the set @xmath304 . using and , we compute the following error bound @xmath305 , \quad 1 \leq j \leq n,\\ \nonumber & \leq 4 \ , { \mathbb{e } } [ \mathbbm{1 } _ { \ { u_{j } \in a_r^c \ } } |u_{j } |^2 ] \leq 4 \ , ( { \mathbb{e}}|u_j|^p ) ^{2/p } \ , { \mathbb{p}}(u_j \in a_r^c)^{1/q } , \quad p=2+\epsilon , q= 1 + 2\epsilon^{-1 } , \\
\label{eq : ujrbound } & \leq c_{s , n , c_0 } \ , r^{-2/q}.\end{aligned}\ ] ] then , using property [ item : one ] gives @xmath306 taking expectation with respect to remaining measures and using yields @xmath307 setting @xmath308 in entails @xmath309 since @xmath310 , we also have @xmath311 .
thus , we can choose @xmath312 large enough such that @xmath313 for each @xmath296 .
based on the preceding lemma , we prove that the chain converges to the solution of the chain in the setting of a finite number of restarts .
[ thm : convfinite ] under the assumptions of lemma [ lemma : rapprox ] , for each @xmath56 , there exists @xmath314 and then @xmath315 such that @xmath316 for each @xmath296 .
let @xmath317 be fixed . by the previous lemma ,
choose @xmath276 such that @xmath318 for @xmath296 .
then , using calculations similar to those above , by [ item : one ] we get @xmath319 where @xmath320 and @xmath321 .
then by [ item : two ] , we choose @xmath322 sufficiently large so that @xmath323 the last inequality can be rewritten as @xmath324 therefore , @xmath325 for each @xmath326 . consider the setting of the solution operator to the sde given in .
as @xmath293 increases to @xmath327 , the random variable @xmath328 may converge in distribution to a limiting random variable @xmath329 , whose distribution is the invariant measure of the evolution equation .
although there are many efficient ways to analyze and compute such invariant measures ( see for instance @xcite ) , we wish to show that our iterative algorithm also converges to that invariant measure as degrees of freedom tend to infinity . in other words , our pce - based method allows us to remain accurate for long - time evolutions something that can not be achieved without restarts .
now we iterate each discrete chain in , and for an arbitrary @xmath330 . in order to ensure long - time convergence , we need a stricter condition than [ item : one ] and impose instead the following contraction condition 1 .
@xmath273 , where @xmath331 , @xmath332 @xmath272 , and @xmath274 .
[ item : onemod ] we first prove the following result about the existence and uniqueness of an invariant measure for the original chain .
[ lemma : invariant ] under conditions [ item : zero ] and [ item : onemod ] , there exists a unique invariant measure @xmath333 of the chain with bounded @xmath280th moment .
moreover , if @xmath310 , then @xmath334 is bounded uniformly in @xmath83 . for @xmath335 ( and using that @xmath336 ) , we compute @xmath337 where @xmath338 is chosen so that @xmath339 and @xmath340
. thus , there exists a lyapunov function @xmath341 with a constant @xmath342 .
this in turn implies the existence of an invariant measure @xmath333 with finite @xmath280th moment for the process ; see ( * ? ? ?
* corollary 4.23 ) .
uniqueness also follows from assumption [ item : onemod ] ; see ( * ? ? ?
* theorem 4.25 ) .
let @xmath343 be a random variable with law @xmath333 and independent of @xmath46 .
consider the chain @xmath344 started from @xmath343 .
note @xmath345 for all @xmath330 .
then , from the bound @xmath346 we conclude that @xmath347 .
the following theorem establishes the exponential convergence of the pc chain to the chain as time @xmath83 increases .
assume [ item : zero ] , [ item : onemod ] and [ item : two ] hold , and @xmath310 .
then , for each @xmath56 , there exists @xmath276 independent of @xmath83 , and then @xmath348 , such that @xmath316 for all @xmath349 .
let @xmath317 be given . by lemma [ lemma : invariant ]
, @xmath334 is bounded uniformly in @xmath83 .
the inequality holds with a constant independent of @xmath83 .
then , choosing @xmath350 so that @xmath351 in implies that there exists a constant @xmath352 independent of @xmath83 such that @xmath353 . as in the proof of theorem [ thm : convfinite ] , we choose @xmath350 so that @xmath354
. then , with @xmath348 large enough , is replaced by @xmath355 therefore , @xmath356 this concludes our proof of convergence of the dgpc method to the invariant measure @xmath333 .
let us conclude this section with the following remark
. it would be desirable to obtain that the chain ( corresponding to @xmath357 , or equivalently no support truncation ) @xmath358 remains determinate .
we were not able to do so and instead based our theoretical results on the assumption that the true distributions of interest were well approximated by compactly supported distributions . in practice , the range of @xmath159 has to remain relatively limited for at least two reasons .
first , large @xmath159 rapidly involves very large computational costs ; and second , the determination of measures from moments becomes exponentially ill - posed as the degree of polynomials @xmath25 increases .
for these reasons , the support truncation has been neglected in the following numerical section since , heuristically , for large @xmath359 and limited @xmath25 , the computation of ( low order ) moments of distributions with rapidly decaying density at infinity is hardly affected by such a support truncation .
in this section , we present several numerical simulations of our algorithm and discuss the implementation details .
we consider several of the equations described in @xcite to show that pce - based simulations may perform well in such settings .
we mostly consider the following two - dimensional nonlinear system of coupled sdes : @xmath360 where @xmath361 , @xmath362 are damping parameters , @xmath363 are constants , and @xmath364 and @xmath365 are two real independent wiener processes . the system was proposed in @xcite to study filtering of turbulent signals which exhibit intermittent instabilities .
the performance of hermite pce is analyzed in various dynamical regimes by the authors in @xcite , who conclude that truncated pce struggle to accurately capture the statistics of the solution in the long term due to both the truncated expansion of the white noise and neglecting higher order terms , which become crucial because of the nonlinearities . for a review of the different dynamical regimes that exhibits ,
we refer the reader to @xcite . for the rest of the section
, @xmath366 stands for the endpoint of the time interval while @xmath367 denotes the time - step after which restarts occur at @xmath368 .
moreover , @xmath82 denotes the number of basic random variables @xmath27 used in the truncation of the expansion . in the presence of two wiener processes
it will denote the total number of variables .
we slightly change the previous notation and let @xmath25 and @xmath369 denote the maximum degree of the polynomials in @xmath14 and @xmath74 , respectively . recall that @xmath74 is the projected pc solution at @xmath50 .
furthermore , following @xcite we choose the orthonormal bases for @xmath370 $ ] as @xmath371.\end{aligned}\ ] ] other possible options for @xmath372 include sine functions , a combination of sines and cosines , and wavelets .
we refer the reader to @xcite for details on the use of wavelets to deal with discontinuities for random inputs . assuming that the solution of is square integrable
, we utilize intrusive galerkin projections in order to establish the following equations for the coefficients of the pce : @xmath373 , \\
\dot{v}_{\alpha } & = -b_v\ , v_{\alpha}+ a_v \ , ( uv)_{\alpha } + \sigma_v \,e[\dot{w}_v t_{\alpha}].\end{aligned}\ ] ] this system of odes is then solved by either a second- or a fourth - order time integration method .
finally , we note that other methods are also available to compute the pce coefficients such as nonintrusive projection and collocation methods @xcite . in most of our numerical examples ,
the dynamics converge to an invariant measure .
to demonstrate the convergence behavior of our algorithm , we compare our results to exact second order statistics or monte carlo simulations with sufficiently high sampling rate @xmath374 ( e.g. , euler maruyama or weak runge
kutta methods @xcite ) where the exact solution is not available .
comparisons involve the following relative pointwise errors : @xmath375 where @xmath376 represents the mean and variance . in the following figures , we plot the evolution of mean , variance , @xmath377 and @xmath378 using the same legend .
in addition , we exhibit the evolution of higher order cumulants , which will be denoted by @xmath379 to demonstrate the convergence to steady state as time grows . finally ,
in our numerical computations , we make use of the c++ library uq toolkit @xcite . as a first example , we consider the one - dimensional ornstein - uhlenbeck ( ou ) process @xmath380 , \quad v(0 ) = v_0,\end{aligned}\ ] ] on @xmath381 $ ] with the parameters @xmath382 .
we first present the results of hermite pce .
figure [ fig : ex1_1 ] shows that hermite pc captures the mean accurately but approximation for the variance is accurate only at the endpoint .
this is a consequence of the expansion in , which violates the markov property of brownian motion and is inaccurate for all @xmath383 , while it becomes ( spectrally ) accurate at the endpoint @xmath40 . using frequent restarts , these oscillations
are significantly attenuated as expected ; see figure [ fig : ex1_2b ] .
this oscillatory behavior could also be alleviated by expanding the brownian motion as in on subintervals of @xmath384 .
note that the global basis functions @xmath385 may also be filtered by taking the conditional expectation @xmath386 $ ] , where @xmath387 is the @xmath3-algebra corresponding to brownian motion up to time @xmath293 @xcite .
we do not pursue this issue here but note that the accuracy of the different methods should be compared at the end of the intervals of discretization of brownian motion .
next , we demonstrate numerical results for dgpc taking @xmath206 , @xmath388 and a varying @xmath389 .
figure [ fig : ex1_2 ] shows that as the exact solution converges to a steady state , our algorithm captures the second order statistics accurately even with a small degrees of freedom utilized in each subinterval .
moreover , although we do not show it here , it is possible , by increasing @xmath369 , to approximate the higher order zero cumulants @xmath390 , with an error on the order of machine precision , which implies that the algorithm converges numerically to a gaussian invariant measure . in figure
[ fig : ex1_3 ] , we illustrate an algebraic convergence of the variance in terms of the dimension @xmath82 for @xmath391 and @xmath392 .
dgpc uses the same size of interval @xmath393 for both cases . for comparison
, we also include the convergence behavior of hermite pce .
values of @xmath82 are taken as @xmath394 for figures [ fig : ex1_3a ] and [ fig : ex1_3c ] and @xmath395 for figures [ fig : ex1_3b ] and [ fig : ex1_3d ] .
we deduce that our algorithm maintains an algebraic convergence of order @xmath396 for both @xmath391 and @xmath392 whereas the convergence behavior of hermite pce drops from @xmath396 to @xmath397 as time increases .
this confirms the fact that the degrees of freedom required for standard pce to maintain a desired accuracy should increase with time .
we now introduce a nonlinearity in the equation so that the damping term includes a cubic component : @xmath398 figure [ fig : ex2_1 ] displays several numerical simulations for the degree of polynomials in @xmath14 given by @xmath399 and @xmath400 .
this is compared with the weak runge
kutta method using @xmath401 and @xmath402 .
we observe that increasing @xmath25 improves the accuracy of the solution to this nonlinear equation as expected .
moreover , table [ table : cubic ] presents a comparison for statistical cumulants between our method with @xmath403 , @xmath404 , @xmath405 and monte carlo ( mc ) at time @xmath406 .
the ( stationary ) cumulants of the invariant measure are also estimated by solving a standard fokker
planck ( fp ) equation @xcite for the invariant measure .
we conclude this example by noting that the level of accuracy of approximations in dgpc for cumulants is similar to that of the mc method .
|r | r | r |r|r|r|r| & @xmath407&@xmath408 & @xmath409 & @xmath410&@xmath411 & @xmath412 + dgpc & 3.58e-5 & 7.33e-1 & -2.08e-5 & -3.37e-1 & 6.02e-5 & 9.35e-1 + mc & -2.53e-3 & 7.33e-1 & 7.85e-4 & -3.38e-1 & -3.34e-3 & 9.58e-1 + fp & 0 & 7.33e-1 & 0 & -3.39e-1 & 0 & 9.64e-1 + as a third example we consider an ou process in which the damping parameter is random and uniformly distributed in @xmath413 $ ] , i.e. @xmath414 .
this is an example of non - gaussian dynamics that may be seen as a coupled system for @xmath415 with @xmath416 .
we consider a time domain @xmath417 $ ] and divide it into @xmath418 subintervals .
the initial condition is normally distributed @xmath419{90}{$\models$}}w_v$ ] and @xmath400 . in the next figure
, we compare second order statistics obtained by our method to monte carlo method for which we use the euler
maruyama scheme with the time step @xmath420 and sampling rate @xmath421 implying @xmath422 samples in total .
we stress again that this problem is essentially two - dimensional since the damping is random .
as expected , the mean decreases monotonically and is approximated accurately by the algorithm .
the estimation of the variance becomes more accurate as @xmath25 increases .
furthermore , figures [ fig : ex3_1c ] and [ fig : ex3_1d ] show that the cumulants become stationary for long times indicating that the numerical approximations converge to a measure which is non - gaussian .
table [ table : ex3 ] compares the first six cumulants obtained by our algorithm at time @xmath423 with @xmath424 , @xmath425 , @xmath405 to the monte carlo method . for further comparison
, we also provide cumulants obtained by averaging fokker
planck density with respect to the known , explicit , distribution of the damping .
it can be observed from the following table that both our algorithm and monte carlo capture cumulants reasonably well although the accuracy degrades at higher orders .
|r | r | r |r|r|r|r| & @xmath407&@xmath408 & @xmath409 & @xmath410&@xmath411 & @xmath412 + dgpc & 1.68e-5 & 1.10 & 3.80e-5 & 3.78e-1 & 8.72e-5 & 5.04e-1 + mc & -1.78e-4 & 1.10 & -3.25e-3 & 3.70e-1 & -6.34e-2 & 4.65e-1 + fp & 0 & 1.10 & 0 & 3.79e-1&0 & 5.29e-1 + the above calculations provide an example of stochasticity with two distinct components : the variables @xmath42 that are markov and can be projected out and the non - markov variable @xmath426 that has long - time effects on the dynamics . the variables @xmath415 are strongly correlated for positive times .
pce thus need to involve orthogonal polynomials that reflect these correlations and can not be written as tensorized products of orthogonal polynomials in @xmath98 and @xmath426 .
we demonstrate that our method is not limited to equations forced by brownian motion .
we apply it to an equation where the forcing is a nonlinear function of brownian motion : @xmath427 with parameters @xmath428 , @xmath429 . to depict the effect of different choices of number of restarts , we take @xmath430 using @xmath403 , @xmath404 , @xmath431 in each expansion . from figure
[ fig : ex4 ] , we observe that second order statistics are captured accurately and an increasing number of restarts provides better approximations as anticipated .
although the system does not converge to a steady state ( variance of @xmath432 increases linearly ) , this example illustrates that dgpc is able to capture behaviors of solutions where the forcing is a nonlinear function of wiener process .
this example concerns the nonlinear system , where the dynamics of @xmath20 exhibit intermittent non - gaussian behavior . a time dependent deterministic periodic forcing is considered with the second equation being an ou process , i.e. @xmath433 in , and the parameters are taken as @xmath434 , @xmath435 , @xmath436 , @xmath437 , @xmath438 , @xmath439 , with initial conditions @xmath440{90}{$\models$}}v_0 = n(0 , { \sigma_v^2/8b_v } ) $ ] ; see also @xcite .
the initial variables are independent of each other and of the stochastic forcing .
intermittency is introduced by using an ou process , which acts as a multiplicative noise fluctuating in the damping .
the second order statistics for this case ( @xmath433 ) can be derived analytically @xcite and we plot them in black in the following figures using quadrature methods with sufficiently high number of quadrature points .
we first depict the results for @xmath20 obtained by standard hermite pce in the first row of figure [ fig : ex6_1 ] .
it can be observed that truncated pc approximations are accurate for short times .
however , they quickly lose their accuracy as time grows .
this is consistent with the simulations presented in @xcite .
the second row of the same figure shows that the accuracy is vastly improved in dgpc using @xmath441 , and second order statistics are accurate up to three or four digits .
moreover , errors in the variance oscillate around @xmath442 for @xmath443 throughout the time length suggesting that the approximation retains its accuracy in the long term , which is consistent with theoretical predictions .
our algorithm easily outperforms standard pce in capturing the long - time behavior of the nonlinear coupled system in this scenario .
+ for this scenario , we consider the system parameters as @xmath444 , @xmath433 , @xmath445 , @xmath446,@xmath447 , @xmath448 without the deterministic forcing @xmath449 and with the initials @xmath450{90}{$\models$}}\ , v_0 = n(0,{\sigma_v^2/8b_v } ) .
$ ] in this regime , the dynamics of @xmath328 are characterized by unstable bursts of large amplitude @xcite .
second order statistics obtained by hermite pce and our algorithm are presented in figure [ fig : ex7_2 ] , where the first two parts ( figures [ fig : ex7_1a ] and [ fig : ex7_1b ] ) correspond to hermite pce .
we observe from figure [ fig : ex7_1b ] that the relative errors for standard hermite expansions are unacceptably high . on the other hand , as the degrees of freedom is increased in dgpc , there is a clear pattern of convergence of second order statistics ; see figures [ fig : ex7_2e ] and [ fig : ex7_2f ] .
note , however , that short - time accuracy is better than accuracy for long times .
this behavior may be explained by the onset of intermittency as nonlinear effects kick in after short times .
further , figure [ fig : ex7_2 g ] shows the error behavior of variance at @xmath451 as @xmath25 and @xmath369 vary , and @xmath82 is fixed .
the rate of convergence is consistent with exponential convergence in @xmath25 and @xmath369 as the logarithmic plot is almost linear . for similar convergence behaviors ,
see @xcite . + + finally , in figure [ fig : ex7_3 ] , we present the evolution of the first few cumulants obtained by dgpc , which shows that the system converges to a steady state distribution as time increases .
bivariate cumulants are ordered so that first and second subscripts correspond to @xmath20 and @xmath98 , respectively .
figure [ fig : ex7_3f ] shows kurtosis excess for @xmath452 and clearly indicates that the dynamics of @xmath98 stay gaussian , whereas those of @xmath20 converge to a non - gaussian state . as a final example , using the same set of parameters of the previous example , we introduce a small perturbation to the second equation and set @xmath453 in .
another choice of perturbation as @xmath454 was considered before in @xcite . comparing figurs [ fig : ex7_3 ] and [ fig : ex8 ]
, we see that evolutions of higher order cumulants are perturbed in most cases as expected . in this case
, it took a longer transient time to converge to an invariant measure as additional nonlinearity was introduced into the system .
polynomial chaos expansions ( pce ) applied to the simulation of evolution equations suffer from two drawbacks : in the presence of stochastic forcing , the dimension of randomness is too large , and the long - time solution may not be sparsely represented in a fixed pce basis @xcite . in the setting of markovian random forcing , we have proposed a restart method that addresses the two aforementioned drawbacks .
such restarts of the pce , which we called here dynamical generalized polynomial chaos ( dgpc ) , allow us to both to keep the number of random variables small and to obtain a solution that remains reasonably sparse in the evolving basis of orthogonal polynomials , as was done earlier in @xcite . following @xcite , we applied dgpc to the long - time numerical simulation of various stochastic differential equations ( sdes ) . to demonstrate the ability of the algorithm to reach long - time solutions , we computed invariant measures for sdes that admit one and found a very good agreement between dgpc and other standard methods such as monte carlo - based simulations .
we also presented a simple theoretical justification for such a convergence .
the main computational difficulty in dgpc , as in most gpc - based methods , is the estimation of the orthogonal polynomials of evolving arbitrary measures .
our method , as in @xcite , is based on estimating moments of the multivariate distribution of interest and estimating orthogonal polynomials by a gram
schmidt procedure .
the bottleneck in such estimations is the cost of computing moments .
this is similar to the cost of solving differential equations by pce methods , where moment estimations are also the most costly step .
however , in dgpc , such estimations can not be performed offline . from a theoretical point of view
, we need to ensure that the evolving measures remain _ determinate _ so that their orthogonal polynomials span square integrable functionals . since distributions with compact support are determinate , our theoretical results have been applied in the setting where sde solutions are well approximated by compactly supported distributions . in this connection
, we note that the calculation of orthogonal polynomials from knowledge of moments is an ill - posed problem , which is a serious impediment to gpc methods in general .
extension of the method to larger systems than those considered here , including to systems obtained by solving stochastic partial differential equations ( spdes ) is challenging .
the method works reasonably well for low dimensional sdes and it needs modifications for larger systems .
however , the general idea remains applicable . to keep the number of uncertain variables in check
, the method has to be coupled with a compression technique such as karhunen
loeve projection at each restart time .
moreover , sparse truncation techniques further help to reduce costs .
the analysis of extensions of dgpc to larger systems , faster methods to compute orthogonal polynomials , and adaptive restart procedures is the subject of ongoing research .
this work was partially funded by afosr grant nsseff- fa9550 - 10 - 1 - 0194 and nsf grant dms-1408867 .
height 2pt depth -1.6pt width 23pt , _ reduced chaos expansions with random coefficients in reduced - dimensional stochastic modeling of coupled problems _ , internat .
methods engrg . , 97 ( 2014 ) ,
352376 .
, _ moment problems related to the solutions of stochastic differential equations _ , in stochastic theory and control ( lawrence , ks , 2001 ) , vol . 280 of lecture notes in control and inform .
sci . , springer , berlin , 2002 , pp . | polynomial chaos expansions ( pce ) allow us to propagate uncertainties in the coefficients of differential equations to the statistics of their solutions .
their main advantage is that they replace stochastic equations by systems of deterministic equations .
their main challenge is that the computational cost becomes prohibitive when the dimension of the parameters modeling the stochasticity is even moderately large .
we propose a generalization of the pce framework that allows us to keep this dimension as small as possible in favorable situations .
for instance , in the setting of stochastic differential equations ( sdes ) with markov random forcing , we expect the future evolution to depend on the present solution and the future stochastic variables .
we present a restart procedure that precisely allows pce to depend only on that information .
the computational difficulty then becomes the construction of orthogonal polynomials for dynamically evolving measures .
we present theoretical results of convergence for our dynamical generalized polynomial chaos ( dgpc ) method .
numerical simulations for linear and nonlinear sdes show that it adequately captures the long - time behavior of their solutions as well as their invariant measures when the latter exist .
keywords : polynomial chaos , stochastic differential equations , uncertainty quantification , markov processes |
supernova 2001du was visually discovered by r. evans ( 2001 ) on august 24.7 ( ut dates are used throughout this paper ) at the western end of the bar in the spiral galaxy ngc 1365 ( about 90 w and 10 s of the nucleus ) .
smartt , kilkenny , & meikle ( 2001 ) classified the sn as type ii ; wang et al . ( 2001 ) , from a spectrum obtained on august 30 ut , confirmed the sn as type ii - plateau ( ii - p ) before optical maximum .
ngc 1365 has also been host to sn 1957c ( of unknown type ) and the type ic sn 1983v . using
_ hubble space telescope _
( _ hst _ ) observations of cepheids , silbermann et al .
( 1999 ) determined the distance modulus of ngc 1365 to be @xmath1 mag , corresponding to a distance of about 18.2 mpc .
van dyk , li , & filippenko ( 2003 ) recently measured the absolute position of sn 2001du and , applying an independent astrometric grid to pre - sn _ hst _ archival wfpc2 images , isolated three possible candidate progenitor stars within the uncertainty of the measured sn position : one reddish star with a detection in both the f555w and f814w bands , and two stars with only f555w detections , implying bluer colors for these latter two stars . also , multi - band wfpc2 images of the sn were obtained for _ hst _ program go-9041 ( pi : s. smartt ) in 2001 november .
these images are now in the public domain , and in this paper we analyze them and compare them to the pre - sn images to further constrain the nature of the sn progenitor .
( smartt et al .
2003 have also recently analyzed their images and , in a somewhat similar manner , come to the same basic conclusion that we reach here . )
pre - sn images , each of total exposure time 300 s in the bands f336w , f555w , and f814w , were obtained by gto-5222 on 1995 january 15 .
late - time multi - band images of the sn were obtained by go-9041 on 2001 november 26 .
these include deep images in f336w , f439w , f555w , f675w , and f814w , as well as shorter exposures in f336w , f555w , and f814w .
the available _ hst _ data are summarized in table 1 , with total exposure times given for each band . in van dyk et al .
( 2003 ) we adopted the position of the sn , measured from an image obtained with the yalo 1-m telescope at ctio ( and available on the internet ) , as @xmath2 = 3@xmath333@xmath429@xmath515 , @xmath6 = @xmath7360831@xmath85 ( j2000 ) , with total uncertainty @xmath9 . the sn site is located on the wf3 chip . in figure 1
we reproduce figure 21 from van dyk et al .
, showing the sn environment in the f555w band . in figure 2
we show the late - time f555w image ( the sn is nearly centered on the pc chip for all the go-9041 exposures ) .
as one can see , the actual sn position is within the @xmath10 error circle , which provides additional confidence in the astrometric method used by van dyk et al . when isolating the various sn sites = 3@xmath333@xmath429@xmath514 , @xmath6 = @xmath7360831@xmath83 ( j2000 ) , @xmath11 in each coordinate .
we initially used the routine hstphot ( dolphin 2000a , b ) to obtain photometry of the objects imaged on the pc chip .
hstphot automatically accounts for wfpc2 point - spread function ( psf ) variations and charge - transfer effects across the chips , zeropoints , aperture corrections , etc . , and can return as output magnitudes transformed from the _ hst _ flight system to standard johnson - cousins bands .
van dyk et al . ( 2003 ) discuss the similarity in the results from hstphot , compared with those from the packages dophot ( schechter , mateo , & saha 1993 ) and daophot / allstar ( stetson 1987 , 1992 ) . and @xmath12 mag . ] the sn profile
is saturated in the longer f555w ( @xmath13 ) , f675w ( @xmath14 ) , and f814w ( @xmath15 ) exposures .
thus , no brightness measurement is possible .
the sn is not saturated in the shorter f555w and f814w exposures , the long f336w ( @xmath16 ) and f439w ( @xmath17 ) exposures , and also the short f336w exposure
. however , the results from hstphot for the f336w exposures varied , depending on how these exposures were combined with the other bands as input into the routine , and the results from the long and short exposures for this band were also discrepant with each other ( the results ranged from about @xmath18 to 18.7 mag for different runs of hstphot ) .
we do not know why this occurred , but we speculate that it possibly may be due to the low signal - to - noise ratio for both exposure sets . for f336w
we then applied to both the long and short exposures daophot / allstar within iraf as a check , using a tinytim psf ( krist 1995 ) , the zeropoint from holtzman et al .
( 1995 ) , and aperture corrections ( @xmath190.00 mag ) derived from artificial star tests .
( checks for the f439w and short f555w and f814w exposures with daophot show good agreement with the hstphot results . )
we obtained @xmath20 and 18.87@xmath21 mag for the short and long exposures , respectively .
we then averaged these daophot magnitudes with the hstphot magnitudes at f336w , and obtained @xmath22 mag .
the sn was quite red by late november 2001 , with @xmath23 mag .
whether using the transformations from flight system to johnson - cousins magnitudes from holtzman et al .
( 1995 ) , dolphin ( 2000b ) , or via synphot applied to the bruzual spectral synthetic atlas ( see van dyk , filippenko , & li 2002 ) , this color is too red to obtain a reliable transformation from @xmath24 to @xmath16 .
however , we do apply a transformation , with uncertainty @xmath190.2 mag .
the transformations for all other bands are from dolphin ( 2000b ) , applied internally within hstphot .
the final results for the photometry of sn 2001du in the _ hst _ images are presented in table 2 . in figure 3
we show preliminary , unpublished @xmath25 light curves for sn 2001du , based on data from the yalo 1-m telescope ( n. suntzeff & k. krisciunas , 2003 , private communication ) , and we include the @xmath26 photometry from the _ hst _ data . for comparison
we show the @xmath17 and @xmath13 light curves for the plateau - type sn ( ii - p ) 1999em in ngc 1637 ( leonard et al .
it is clearly evident that sn 2001du is indeed a sn ii - p as well .
if this sn were a linear - type event ( ii - l ) , it should have markedly declined in brightness by the time of the _ hst _ observations ( @xmath1994 days since discovery ) .
in van dyk et al . ( 2003 ) we had isolated three faint objects , a c , detected and determined to be stellar by hstphot , within the error circle around the sn position . stars a and b , to the south and west , respectively , are both blue , with @xmath27 ( @xmath28 ) @xmath29 mag and @xmath30 mag , respectively , and no detections at f814w ( @xmath31 ) .
( blue , here , is relative , and includes stars with roughly early - g spectral type or earlier , based on the detection limits ; see below . )
star c , to the east , is relatively red , with @xmath32 and @xmath33 mag .
we had assumed that , since sn 2001du is of type ii - p , the most plausible of the three candidates is the redder star . however , that was argued to be still unlikely , based on the assumed luminosity and color of star c. we have used the late - time sn images , specifically the @xmath13 image , and attempted to match them up with the pre - sn images .
this was done by using the task _ geotran _ in iraf to transform both the late - time sn images to the pre - sn images and vice versa .
we can therefore locate the position of the sn on the pre - sn images as being [email protected] wf pixel ( @xmath35 ) northeast of the position of star b. it would be tempting to assign star b , then , as the progenitor of sn 2001du . however , ( 1 ) the offset of the sn position from star b ( which corresponds spatially to 6.2 pc at the distance of ngc 1365 ) is significantly large , and the accuracy of the measurement is appreciably high ; ( 2 ) sne ii - p are thought to arise from the explosions of red supergiants ; and ( 3 ) no evidence exists that sn 2001du is peculiar in any way and that it arose from a bluer , more compact star like the progenitor of sn 1987a ( e. cappellaro , 2002 , private communication ) .
thus , it is more likely , as we previously had suspected ( van dyk et al .
2003 ) , that the progenitor simply is not detected in the pre - sn images .
we can use the detection limits of both the f555w and f814w pre - sn images to constrain the nature of the progenitor .
we do this by inferring limits on the luminosity and , therefore , the mass from the limits on the star s apparent magnitude .
these hstphot detection limits in the sn environment are @xmath36 and @xmath37 mag . before converting into absolute magnitude
, we must first assess the amount of extinction to the sn . in van dyk et al .
( 2003 ) we attempted to use stars in the environment to estimate the reddening .
the galactic extinction toward ngc 1365 is @xmath38 mag ( schlegel , finkbeiner , & davis 1998 ; and ned ) .
we had found that stars in the larger sn environment implied larger amounts of extinction , @xmath39 mag .
( unfortunately , although the post - sn images are deeper than the pre - sn ones , the sn is far too bright , and its image is spread over too many surrounding pixels , to perform a more thorough study of the environment . )
using the light curve in figure 3 we can actually derive more specific information about the possible reddening toward the sn from the sn itself ( which is information that was not available to us previously ) .
the _ hst _
photometry of the sn on november 26 is about 94 days after discovery , i.e. , near the end of the light - curve plateau , where we can still estimate the sn s color temperature with reasonable accuracy ( e.g. , hamuy et al .
these colors are @xmath40 and @xmath41 mag .
for comparison , the colors of the very well - studied type ii - p sn 1999em at the same age ( about 95 days after discovery ) are @xmath42 and @xmath43 mag ( leonard et al .
leonard et al . and hamuy et al . ( 2001 ) conclude that sn 1999em is reddened by @xmath44 mag . since @xmath45 is affected by line blanketing in the @xmath17 band ( e.g. , hamuy et al .
2001 ) , the small difference in @xmath45 ( sn 1999em is 0.05 mag redder than sn 2001du ) could be due to differences in metallicity between the two sne , which might affect this color .
the marginally redder @xmath46 color of sn 2001du ( by 0.02 mag ) may imply that this sn is slightly more reddened than sn 1999em .
however , we consider these differences in color between the two sne to be small and likely within the uncertainties in the colors , so we adopt @xmath47 mag for sn 2001du , as found for sn 1999em .
if we assume the cardelli , clayton , & mathis ( 1989 ) reddening law , then @xmath48 mag and @xmath49 mag for sn 2001du . for the distance modulus @xmath50 mag , @xmath51 and @xmath52 mag for the progenitor star . in a similar vein to the estimation of the progenitor mass limit for sn ii - p 1999em by smartt et al .
( 2002 ) and sn ii - p 1999gi by smartt et al .
( 2001 ) , the absolute magnitude limits can be converted to the luminosity of the likely supergiant progenitor , assuming the full range of possible stellar surface temperatures .
this is accomplished using a similar relation to equation ( 1 ) in smartt et al .
( 2001 , 2002 ) , where we also have assumed the temperatures and stellar bolometric corrections for supergiants from drilling & landolt ( 2000 ) and have computed , using synphot within iraf and the bruzual spectral atlas , the conversions of @xmath27 and @xmath53 to johnson @xmath13 for the range in spectral type .
the luminosity limits for both f555w ( @xmath13 ) and f814w ( @xmath15 ) are shown in figure 4 .
stars with luminosities brighter than these limits ( shown by the _ heavy lines _ ) should have been detected in the pre - sn images . for comparison
we have plotted ( also in a similar fashion to smartt et al . ) the model stellar evolutionary tracks for a range of masses from lejeune & schaerer ( 2001 ) , assuming enhanced mass loss for the most massive stars and that a metallicity @xmath54 is appropriate for the sn environment .
indication for a metallicity somewhat greater than solar in this sector of ngc 1365 comes from the abundance study by roy & walsh ( 1997 ) of h ii regions in the galaxy . the closest h ii regions to the sn site are their nos . 20 and 21 , where at this radial distance from the nucleus , the o / h abundance is @xmath199 dex or , in fact , slightly greater , based on model assumptions ( the solar o / h abundance is 8.8 dex ; grevesse & sauval 1998 ) .
the luminosity limits at each band in figure 4 show that the pre - sn images are not sensitive to the bluest high - mass supergiants , so we can not be absolutely sure that such a star is not the progenitor .
if we exclude this possibility , the @xmath13 band sets an upper limit to the progenitor mass of @xmath55 , although this ranges upward to as high as @xmath56 , depending on the effective temperature .
the @xmath15 band is more sensitive to less - massive red supergiants , ranging downward to @xmath57 ( possibly even less ) at the lowest effective temperatures .
these results from figure 4 are roughly consistent with what we showed in figure 22 of van dyk et al .
we therefore estimate that the upper mass limit for the sn 2001du progenitor is @xmath0 .
this limit is consistent with the limit leonard et al .
( 2002b ) find for sn ii - p 1999gi ( @xmath58 ) and the @xmath59 limit one finds for sn ii - p 1999em , if the distance to the sn 1999em host is assumed to be the cepheid distance , @xmath60 mpc ( leonard et al .
low masses for the progenitors of sne ii - p , relative to that indicated for sne
ii - l ( linear ) and sne iin ( narrow ) , are also inferred from their optical properties ( e.g. , schlegel 1996 ; filippenko 1997 ; chugai & utrobin 2000 ) and suggested by the inferred paucity of circumstellar matter from radio and x - ray observations ( e.g. , weiler et al .
1989 ; schlegel 2001 ; pooley et al .
comparing _ hst _ archival wfpc2 images obtained of the type ii - p sn 2001du in ngc 1365 at late times with images obtained before explosion , we have located the position of the sn progenitor star to within 1 wf chip pixel of the candidate star b identified by van dyk et al .
this star , however , is more than likely _ not _ the progenitor .
we use the detection limits of the pre - sn f555w and f814w images to place constraints on the progenitor s nature ; specifically , @xmath61 .
( smartt et al .
2003 find a similar mass limit of @xmath62 . )
this limit is consistent with what has been found and inferred for other sne ii - p .
the proximity of the sn to the blue star b implies that the progenitor may have been part of a small stellar association ( the separation between the progenitor and star b is likely too large for the blue star to be a binary companion ) . the faint emission underlying stars a , b , and c identified by van dyk et al .
( 2003 ) suggests that an association complex may lie in the sn environment
. it would be interesting to image the sn again in multiple bands with _ hst _ at high spatial resolution , presumably with the advanced camera for surveys ( acs ) , when the sn has significantly dimmed .
the stars in the environment of the fading sn would then become more obvious , and the age and mass of the progenitor star might be further constrained , based on the characteristics of its surviving neighbors .
this research made use of the nasa / ipac extragalactic database ( ned ) which is operated by the jet propulsion laboratory , california institute of technology , under contract with nasa , and the leda database ( http://leda.univ-lyon.fr ) .
the work of a.v.f.s group at uc berkeley is supported by nasa grants ar-8754 , ar-9529 , and go-8602 from the space telescope science institute , which is operated by aura , inc .
, under nasa contract nas5 - 26555 .
we thank d. c. leonard and m. hamuy for useful discussions .
we also thank the referee , k. krisciunas , for helpful comments and for providing unpublished ground - based photometry of sn 2001du .
lcclc 1995 jan 15 & f336w & 300 & gto-5222 & & f555w & 300 & & & f814w & 300 & & 2001 nov 26 & f336w & 100 & go-9041 & n & & 460 & & n & f439w & 400 & & n & f555w & 40 & & n & & 600 & & y & f675w & 400 & & y & f814w & 40 & & n & & 600 & & y | _ hubble space telescope _ ( _ hst _ ) wfpc2 archival images obtained years prior to the explosion of the type ii - plateau supernova ( sn ) 2001du in ngc 1365 previously have been analyzed to isolate the progenitor star .
the sn site was located using precise astrometry applied to the _ hst _ images , with significant uncertainty , leaving three possible progenitor candidates .
images of the fading sn have recently become publicly available in the _ hst _ archive , allowing us to pinpoint the sn s exact location on the pre - explosion images .
we show that the sn occurred in very close proximity to one of the blue candidate stars , but we argue that this star is not the actual progenitor . instead
, the progenitor was not detected on the pre - sn images , and we constrain the progenitor s mass to be less than @xmath0 .
this is consistent with previous constraints on the progenitor masses of other type ii - plateau supernovae ( sne ) , suggesting that such sne arise from the iron core collapse of massive stars at the lower extreme of the possible mass range . |
ultra - luminous infrared galaxies ( ulirgs ; @xcite ) have total infrared luminosities @xmath2710@xmath28l@xmath20 and implied star formation rates ( sfr ) @xmath31001000m@xmath20yr@xmath11 .
their low space densities mean that ulirgs contribute @xmath291% of the volume average star formation density at @xmath1=0 . however , the first deep , single - dish bolometer surveys in the 870-@xmath0 m atmospheric window uncovered high number densities of high - redshift sub - mm galaxies ( smgs ) at mjy flux density levels @xcite .
subsequent spectroscopic studies of the radio-/mid - infrared detected subset of the population gave a median redshift of @xmath302.5 @xcite , confirming their ulirg - like luminosities @xcite and demonstrated that ulirgs undergo a 1000-fold increase in space density from @xmath1=0 to @xmath302.5 .
thus , in contrast to the local universe , ulirgs are a non - negligible component of the star - forming population at @xmath3023 ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
these 870-@xmath0m - selected samples remain the best - studied smgs , and links have been proposed between smgs , qsos and the formation phase of massive galaxies at high redshift ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
smgs are therefore a potentially important element in models of galaxy formation .
however , these evolutionary links are still unproven , although it is clear that most luminous smgs lie above the purported `` main sequence '' of star - forming galaxies in the star formation rate
m@xmath31 plane ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
along with corroborating kinematic and morphological evidence @xcite , inevitably it has been argued that smgs can be understood as `` scaled - up '' analogs of local ulirgs which are dominated by merger induced starbursts .
it has also been proposed that lower luminosity smgs overlap with the `` main sequence '' , leading to claims of an apparent dichotomy within this population : i.e. between a true `` starburst '' population and slightly less active galaxies , potentially driven by secular processes @xcite .
theoretical attempts to reproduce basic properties of smgs , such as 870-@xmath0 m number counts , have also led to an equally wide variety of conclusions about the nature and diversity of this population .
for example , early models from @xcite and @xcite include two recipes for star formation ; `` burst '' and `` quiescent '' , with smgs corresponding to the most extreme starburst systems . one key strength of these models is that they are also required to fit the galaxy population at @xmath1=0 , which forces the former to adopt a burst initial mass function ( imf ) biased to high - mass stars ( allowing relatively low - mass galaxies to produce intense starbursts ; @xcite ) .
in contrast , numerical models of @xcite ( see also @xcite and @xcite ) suggest that smgs are dominated by secular bursts in gas - rich disk galaxies , with `` standard '' imfs .
hence theoretical models variously predict the smg population to comprise low - mass merging starbursts ( with unusually low mass - to - light ratios ; e.g. @xcite ) and isolated , gas - rich disk galaxies undergoing secular bursts ( e.g. * ? ? ?
* ; * ? ? ?
this rich array of theoretical options demonstrates the power of smgs to distinguish between the competing galaxy evolution theories .
observations of the most basic properties of smgs , such as their redshift distribution , bolometric luminosities , cold molecular gas dynamics and gas mass , and contribution to the cosmic star formation density should have the power to distinguish between the theoretical models .
however , the coarse resolution of single dish sub - mm telescopes ( typically @xmath321520@xmath33 ) means that identifying the smg counterparts has to date , relied on correlations between their sub - millimetre emission and that in other wave - bands where higher spatial resolution is available ( usually the radio and/or mid - infrared ; e.g. @xcite ) . the spectral energy distributions ( seds ) in these other wave - bands have positive k - corrections , making counterparts which either lie at higher redshift or that have colder - than - average dust temperatures impossible to identify .
indeed , in sub - millimetre surveys typically 4050% of sub - mm sources lack `` robust '' counterparts in the radio or mid - infrared ( e.g. @xcite , see also @xcite ) and it is unknown whether the unidentified smgs have the same redshift distribution ( or are representative ) of the radio - identified subset , potentially biasing the current observational results . to circumvent the problem of missing- ( and mis- ) identifications and so characterize the _ whole _ population of bright smgs in an unbiased manner requires precisely locating the sub - mm emission using sub / mm interferometers ( e.g. * ? ? ?
* ; * ? ? ?
recently , we have undertaken an alma survey of the 126 sub - mm sources in the [email protected] degree extended _ chandra _ deep field south ( ecdfs ) , taken from the `` less '' survey @xcite .
the alma data precisely locate the smgs , _ directly _ pin - pointing the source(s ) responsible for the sub - mm emission ( to within @xmath340.3@xmath33 ) , without recourse to statistical radio / mid - ir associations and so yielding unambiguous identifications for the majority of the smgs .
the first results from our survey include the source catalog and multiplicity of bright smgs @xcite ; the first high - resolution sub - mm counts @xcite ; optical and near - infrared properties , photometric redshift distribution and stellar masses ( simpson et al .
2013 ) ; serendipitous identification [ cii ] at @xmath1=4.4 in two smgs @xcite and subsequent @xmath8co follow - up in these galaxies @xcite ; x - ray properties and agn fraction of alma smgs @xcite ; far - infrared radio correlation of smgs ( thomson et al .
2014 submitted ) and _ hst _ morphologies ( simpson et al .
2014 in prep ) . in this paper
, we extend these analysis and exploit the _
space observatory pacs and spire imaging of the ecdfs to derive the far - infrared / sub - millimeter properties of the smgs pinpointed by alma .
in particular , we use the far - infrared ( 70870@xmath0 m ) and 1.4ghz radio photometry together with new optical / mid - infrared - derived photometric redshifts from simpson et al .
( 2013 ) to derive the far - infrared luminosities , characteristic dust temperatures , dust masses and investigate their evolution with redshift . in
2 we describe the observations and deblending techniques ; in 3 we present the far - infrared properties ( colours , luminosities , dust masses ) of both individual aless smgs and those of the stacks of subsets of population .
we present our main conclusions in 4 . throughout the paper , we adopt a chabrier imf @xcite and use a @xmath35cdm cosmology @xcite with @xmath36=0.73 , @xmath37=0.27 , and @xmath38=72kms@xmath11mpc@xmath11 . | we exploit alma 870-@xmath0 m ( 345ghz ) observations of sub - millimetre sources in the extended _ chandra _ deep field south to investigate the far - infrared properties of high - redshift sub - millimetre galaxies ( smgs ) . using the precisely located 870@xmath0 m alma positions of 99 smgs , together with 24@xmath0 m and radio imaging of this field
, we deblend the _ herschel_/spire imaging of this region to extract their far - infrared fluxes and colours .
the median redshifts for alma less ( aless ) smgs which are detected in at least two spire bands increases as expected with wavelength of the peak in their seds , with @[email protected] , [email protected] and [email protected] for the 250 , 350 and 500-@xmath0 m peakers respectively .
we find that 34 aless smgs do not have a @xmath33@xmath4 counterpart at 250 , 350 or 500@xmath0 m .
these galaxies have a median photometric redshift derived from the rest - frame uv mid - infrared seds of @[email protected] , which is higher than the full aless smg sample ; @[email protected] . using the photometric redshifts together with the 250870@xmath0 m photometry
, we estimate the far - infrared luminosities and characteristic dust temperature of each smg .
the median infrared luminosity and characteristic dust temperature of the @xmath52mjy smgs is @xmath6=([email protected])@xmath710@xmath8@xmath9 ( star formation rate of sfr=300@xmath230@xmath10yr@xmath11 ) and @xmath12=32@xmath21k ( @xmath13=93@xmath24@xmath0 m ) . at a fixed luminosity ,
the characteristic dust temperature of these high - redshift smgs is @xmath14=35k lower than comparably luminous galaxies at @xmath1=0 , reflecting the more extended star formation occurring in these systems . by extrapolating the 870@xmath0 m number counts to @xmath15=1mjy
, we show that the contribution of @xmath161mjy smgs to the cosmic star formation budget is 20% of the total over the redshift range @xmath1714 .
we derive a median dust mass for these galaxies of @xmath18=([email protected])@xmath710@xmath19m@xmath20 and by adopting an appropriate gas - to - dust ratio , we estimate that the typical molecular mass of the aless smgs in our sample is @xmath21=([email protected])@xmath710@xmath22m@xmath20 .
together with the average stellar masses of smgs , @xmath23=(8@xmath21)@xmath710@xmath22m@xmath20 , this suggests an average molecular gas fraction of @xmath2440% .
finally , we use our estimates of the h@xmath25 masses to show that smgs with @xmath51mjy ( @xmath2610@xmath8l@xmath20 ) contain @xmath2410% of the @xmath172 volume - averaged h@xmath25 mass density at this epoch .
galaxies : starburst , galaxies : evolution , galaxies : high - redshift |
Subsets and Splits
Data Mining Abstracts
This query searches for and retrieves 2000 documents containing the phrase 'data mining' in the abstract, providing limited but specific document samples.